| Message ID | 1459303190-20072-6-git-send-email-vishal.l.verma@intel.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Vishal,
[auto build test WARNING on linux-nvdimm/libnvdimm-for-next]
[also build test WARNING on v4.6-rc1 next-20160329]
[cannot apply to xfs/for-next]
[if your patch is applied to the wrong git tree, please drop us a note to help improving the system]
url: https://github.com/0day-ci/linux/commits/Vishal-Verma/dax-handling-of-media-errors/20160330-100409
base: https://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm libnvdimm-for-next
config: x86_64-randconfig-i0-03300245 (attached as .config)
reproduce:
# save the attached .config to linux build tree
make ARCH=x86_64
Note: it may well be a FALSE warning. FWIW you are at least aware of it now.
http://gcc.gnu.org/wiki/Better_Uninitialized_Warnings
All warnings (new ones prefixed by >>):
fs/ext4/indirect.c: In function 'ext4_ind_direct_IO':
>> fs/ext4/indirect.c:719:11: warning: 'ret_saved' may be used uninitialized in this function [-Wmaybe-uninitialized]
ssize_t ret_saved = 0;
^
vim +/ret_saved +719 fs/ext4/indirect.c
703 smp_mb();
704 if (unlikely(ext4_test_inode_state(inode,
705 EXT4_STATE_DIOREAD_LOCK))) {
706 inode_dio_end(inode);
707 goto locked;
708 }
709 if (IS_DAX(inode))
710 ret = dax_do_io(iocb, inode, iter, offset,
711 ext4_dio_get_block, NULL, 0);
712 else
713 ret = __blockdev_direct_IO(iocb, inode,
714 inode->i_sb->s_bdev, iter,
715 offset, ext4_dio_get_block,
716 NULL, NULL, 0);
717 inode_dio_end(inode);
718 } else {
> 719 ssize_t ret_saved = 0;
720
721 locked:
722 if (IS_DAX(inode)) {
723 ret = dax_do_io(iocb, inode, iter, offset,
724 ext4_dio_get_block, NULL, DIO_LOCKING);
725 if (ret == -EIO && iov_iter_rw(iter) == WRITE)
726 ret_saved = ret;
727 else
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
Hi Vishal, still NAK to calling the direct I/O code directly from the dax code.
On Tue, 2016-03-29 at 23:34 -0700, Christoph Hellwig wrote: > Hi Vishal, > > still NAK to calling the direct I/O code directly from the dax code. Hm, I thought this was what you meant -- do the fallback/retry attempts at the callers of dax_do_io instead of the new dax wrapper function.. Did I misunderstand you?
On Wed, Mar 30, 2016 at 12:54:37AM -0600, Vishal Verma wrote: > On Tue, 2016-03-29 at 23:34 -0700, Christoph Hellwig wrote: > > Hi Vishal, > > > > still NAK to calling the direct I/O code directly from the dax code. > > Hm, I thought this was what you meant -- do the fallback/retry attempts > at the callers of dax_do_io instead of the new dax wrapper function.. > Did I misunderstand you? Sorry, it is. I misread fs/block_dev.c as fs/dax.c before my first coffee this morning. I'll properly review the series in the afternoon..
Vishal Verma <vishal.l.verma@intel.com> writes: > dax_do_io (called for read() or write() for a dax file system) may fail > in the presence of bad blocks or media errors. Since we expect that a > write should clear media errors on nvdimms, make dax_do_io fall back to > the direct_IO path, which will send down a bio to the driver, which can > then attempt to clear the error. [snip] > + if (IS_DAX(inode)) { > + ret = dax_do_io(iocb, inode, iter, offset, blkdev_get_block, > NULL, DIO_SKIP_DIO_COUNT); > - return __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, offset, > + if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) > + ret_saved = ret; > + else > + return ret; > + } > + > + ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, offset, > blkdev_get_block, NULL, NULL, > DIO_SKIP_DIO_COUNT); > + if (ret < 0 && ret_saved) > + return ret_saved; > + Hmm, did you just break async DIO? I think you did! :) __blockdev_direct_IO can return -EIOCBQUEUED, and you've now turned that into -EIO. Really, I don't see a reason to save that first -EIO. The same applies to all instances in this patch. Cheers, Jeff > + return ret; > } > > int __sync_blockdev(struct block_device *bdev, int wait) > diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c > index 824f249..64792c6 100644 > --- a/fs/ext2/inode.c > +++ b/fs/ext2/inode.c > @@ -859,14 +859,22 @@ ext2_direct_IO(struct kiocb *iocb, struct iov_iter *iter, loff_t offset) > struct address_space *mapping = file->f_mapping; > struct inode *inode = mapping->host; > size_t count = iov_iter_count(iter); > - ssize_t ret; > + ssize_t ret, ret_saved = 0; > > - if (IS_DAX(inode)) > - ret = dax_do_io(iocb, inode, iter, offset, ext2_get_block, NULL, > - DIO_LOCKING); > - else > - ret = blockdev_direct_IO(iocb, inode, iter, offset, > - ext2_get_block); > + if (IS_DAX(inode)) { > + ret = dax_do_io(iocb, inode, iter, offset, ext2_get_block, > + NULL, DIO_LOCKING | DIO_SKIP_HOLES); > + if (ret == -EIO && iov_iter_rw(iter) == WRITE) > + ret_saved = ret; > + else > + goto out; > + } > + > + ret = blockdev_direct_IO(iocb, inode, iter, offset, ext2_get_block); > + if (ret < 0 && ret_saved) > + ret = ret_saved; > + > + out: > if (ret < 0 && iov_iter_rw(iter) == WRITE) > ext2_write_failed(mapping, offset + count); > return ret; > diff --git a/fs/ext4/indirect.c b/fs/ext4/indirect.c > index 3027fa6..798f341 100644 > --- a/fs/ext4/indirect.c > +++ b/fs/ext4/indirect.c > @@ -716,14 +716,22 @@ retry: > NULL, NULL, 0); > inode_dio_end(inode); > } else { > + ssize_t ret_saved = 0; > + > locked: > - if (IS_DAX(inode)) > + if (IS_DAX(inode)) { > ret = dax_do_io(iocb, inode, iter, offset, > ext4_dio_get_block, NULL, DIO_LOCKING); > - else > - ret = blockdev_direct_IO(iocb, inode, iter, offset, > - ext4_dio_get_block); > - > + if (ret == -EIO && iov_iter_rw(iter) == WRITE) > + ret_saved = ret; > + else > + goto skip_dio; > + } > + ret = blockdev_direct_IO(iocb, inode, iter, offset, > + ext4_get_block); > + if (ret < 0 && ret_saved) > + ret = ret_saved; > +skip_dio: > if (unlikely(iov_iter_rw(iter) == WRITE && ret < 0)) { > loff_t isize = i_size_read(inode); > loff_t end = offset + count; > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c > index dab84a2..27f07c2 100644 > --- a/fs/ext4/inode.c > +++ b/fs/ext4/inode.c > @@ -3341,7 +3341,7 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter, > { > struct file *file = iocb->ki_filp; > struct inode *inode = file->f_mapping->host; > - ssize_t ret; > + ssize_t ret, ret_saved = 0; > size_t count = iov_iter_count(iter); > int overwrite = 0; > get_block_t *get_block_func = NULL; > @@ -3401,15 +3401,22 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter, > #ifdef CONFIG_EXT4_FS_ENCRYPTION > BUG_ON(ext4_encrypted_inode(inode) && S_ISREG(inode->i_mode)); > #endif > - if (IS_DAX(inode)) > + if (IS_DAX(inode)) { > ret = dax_do_io(iocb, inode, iter, offset, get_block_func, > ext4_end_io_dio, dio_flags); > - else > - ret = __blockdev_direct_IO(iocb, inode, > - inode->i_sb->s_bdev, iter, offset, > - get_block_func, > - ext4_end_io_dio, NULL, dio_flags); > + if (ret == -EIO && iov_iter_rw(iter) == WRITE) > + ret_saved = ret; > + else > + goto skip_dio; > + } > > + ret = __blockdev_direct_IO(iocb, inode, > + inode->i_sb->s_bdev, iter, offset, > + get_block_func, > + ext4_end_io_dio, NULL, dio_flags); > + if (ret < 0 && ret_saved) > + ret = ret_saved; > + skip_dio: > if (ret > 0 && !overwrite && ext4_test_inode_state(inode, > EXT4_STATE_DIO_UNWRITTEN)) { > int err; > diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c > index d445a64..7cfcf86 100644 > --- a/fs/xfs/xfs_aops.c > +++ b/fs/xfs/xfs_aops.c > @@ -1413,6 +1413,7 @@ xfs_vm_direct_IO( > dio_iodone_t *endio = NULL; > int flags = 0; > struct block_device *bdev; > + ssize_t ret, ret_saved = 0; > > if (iov_iter_rw(iter) == WRITE) { > endio = xfs_end_io_direct_write; > @@ -1420,13 +1421,22 @@ xfs_vm_direct_IO( > } > > if (IS_DAX(inode)) { > - return dax_do_io(iocb, inode, iter, offset, > + ret = dax_do_io(iocb, inode, iter, offset, > xfs_get_blocks_direct, endio, 0); > + if (ret == -EIO && iov_iter_rw(iter) == WRITE) > + ret_saved = ret; > + else > + return ret; > } > > bdev = xfs_find_bdev_for_inode(inode); > - return __blockdev_direct_IO(iocb, inode, bdev, iter, offset, > + ret = __blockdev_direct_IO(iocb, inode, bdev, iter, offset, > xfs_get_blocks_direct, endio, NULL, flags); > + > + if (ret < 0 && ret_saved) > + ret = ret_saved; > + > + return ret; > } > > /*
On Fri, 2016-04-15 at 12:11 -0400, Jeff Moyer wrote: > Vishal Verma <vishal.l.verma@intel.com> writes: > > > > > dax_do_io (called for read() or write() for a dax file system) may > > fail > > in the presence of bad blocks or media errors. Since we expect that > > a > > write should clear media errors on nvdimms, make dax_do_io fall > > back to > > the direct_IO path, which will send down a bio to the driver, which > > can > > then attempt to clear the error. > [snip] > > > > > + if (IS_DAX(inode)) { > > + ret = dax_do_io(iocb, inode, iter, offset, > > blkdev_get_block, > > NULL, DIO_SKIP_DIO_COUNT); > > - return __blockdev_direct_IO(iocb, inode, I_BDEV(inode), > > iter, offset, > > + if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) > > + ret_saved = ret; > > + else > > + return ret; > > + } > > + > > + ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), > > iter, offset, > > blkdev_get_block, NULL, NULL, > > DIO_SKIP_DIO_COUNT); > > + if (ret < 0 && ret_saved) > > + return ret_saved; > > + > Hmm, did you just break async DIO? I think you did! :) > __blockdev_direct_IO can return -EIOCBQUEUED, and you've now turned > that > into -EIO. Really, I don't see a reason to save that first > -EIO. The > same applies to all instances in this patch. The reason I saved it was if __blockdev_direct_IO fails for some reason, we should return the original cause o the error, which was an EIO.. i.e. we shouldn't be hiding the EIO if the direct_IO fails with something else.. But, how does _EIOCBQUEUED work? Maybe we need an exception for it? Thanks, -Vishal
"Verma, Vishal L" <vishal.l.verma@intel.com> writes: > On Fri, 2016-04-15 at 12:11 -0400, Jeff Moyer wrote: >> Vishal Verma <vishal.l.verma@intel.com> writes: >> > + if (IS_DAX(inode)) { >> > + ret = dax_do_io(iocb, inode, iter, offset, >> > blkdev_get_block, >> > NULL, DIO_SKIP_DIO_COUNT); >> > - return __blockdev_direct_IO(iocb, inode, I_BDEV(inode), >> > iter, offset, >> > + if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) >> > + ret_saved = ret; >> > + else >> > + return ret; >> > + } >> > + >> > + ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), >> > iter, offset, >> > blkdev_get_block, NULL, NULL, >> > DIO_SKIP_DIO_COUNT); >> > + if (ret < 0 && ret_saved) >> > + return ret_saved; >> > + >> Hmm, did you just break async DIO? I think you did! :) >> __blockdev_direct_IO can return -EIOCBQUEUED, and you've now turned >> that >> into -EIO. Really, I don't see a reason to save that first >> -EIO. The >> same applies to all instances in this patch. > > The reason I saved it was if __blockdev_direct_IO fails for some > reason, we should return the original cause o the error, which was an > EIO.. i.e. we shouldn't be hiding the EIO if the direct_IO fails with > something else.. OK. > But, how does _EIOCBQUEUED work? Maybe we need an exception for it? For async direct I/O, only the setup phase of the I/O is performed and then we return to the caller. -EIOCBQUEUED signifies this. You're heading towards code that looks like this: if (IS_DAX(inode)) { ret = dax_do_io(iocb, inode, iter, offset, blkdev_get_block, NULL, DIO_SKIP_DIO_COUNT); if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) ret_saved = ret; else return ret; } ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, offset, blkdev_get_block, NULL, NULL, DIO_SKIP_DIO_COUNT); if (ret < 0 && ret != -EIOCBQUEUED && ret_saved) return ret_saved; There's a lot of special casing here, so you might consider adding comments. Cheers, Jeff
On Fri, 2016-04-15 at 13:11 -0400, Jeff Moyer wrote: > "Verma, Vishal L" <vishal.l.verma@intel.com> writes: > > > > > On Fri, 2016-04-15 at 12:11 -0400, Jeff Moyer wrote: > > > > > > Vishal Verma <vishal.l.verma@intel.com> writes: > > > > > > > > + if (IS_DAX(inode)) { > > > > + ret = dax_do_io(iocb, inode, iter, offset, > > > > blkdev_get_block, > > > > NULL, DIO_SKIP_DIO_COUNT); > > > > - return __blockdev_direct_IO(iocb, inode, > > > > I_BDEV(inode), > > > > iter, offset, > > > > + if (ret == -EIO && (iov_iter_rw(iter) == > > > > WRITE)) > > > > + ret_saved = ret; > > > > + else > > > > + return ret; > > > > + } > > > > + > > > > + ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), > > > > iter, offset, > > > > blkdev_get_block, NULL, > > > > NULL, > > > > DIO_SKIP_DIO_COUNT); > > > > + if (ret < 0 && ret_saved) > > > > + return ret_saved; > > > > + > > > Hmm, did you just break async DIO? I think you did! :) > > > __blockdev_direct_IO can return -EIOCBQUEUED, and you've now > > > turned > > > that > > > into -EIO. Really, I don't see a reason to save that first > > > -EIO. The > > > same applies to all instances in this patch. > > The reason I saved it was if __blockdev_direct_IO fails for some > > reason, we should return the original cause o the error, which was > > an > > EIO.. i.e. we shouldn't be hiding the EIO if the direct_IO fails > > with > > something else.. > OK. > > > > > But, how does _EIOCBQUEUED work? Maybe we need an exception for it? > For async direct I/O, only the setup phase of the I/O is performed > and > then we return to the caller. -EIOCBQUEUED signifies this. > > You're heading towards code that looks like this: > > if (IS_DAX(inode)) { > ret = dax_do_io(iocb, inode, iter, offset, > blkdev_get_block, > NULL, DIO_SKIP_DIO_COUNT); > if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) > ret_saved = ret; > else > return ret; > } > > ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, > offset, > blkdev_get_block, NULL, NULL, > DIO_SKIP_DIO_COUNT); > if (ret < 0 && ret != -EIOCBQUEUED && ret_saved) > return ret_saved; > > There's a lot of special casing here, so you might consider adding > comments. Correct - maybe we should reconsider wrapper-izing this? :) Thanks for the explanation and for catching this. I'll fix it for the next revision. > > Cheers, > Jeff
On Fri, Apr 15, 2016 at 10:37 AM, Verma, Vishal L <vishal.l.verma@intel.com> wrote: > On Fri, 2016-04-15 at 13:11 -0400, Jeff Moyer wrote: [..] >> > >> > But, how does _EIOCBQUEUED work? Maybe we need an exception for it? >> For async direct I/O, only the setup phase of the I/O is performed >> and >> then we return to the caller. -EIOCBQUEUED signifies this. >> >> You're heading towards code that looks like this: >> >> if (IS_DAX(inode)) { >> ret = dax_do_io(iocb, inode, iter, offset, >> blkdev_get_block, >> NULL, DIO_SKIP_DIO_COUNT); >> if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) >> ret_saved = ret; >> else >> return ret; >> } >> >> ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, >> offset, >> blkdev_get_block, NULL, NULL, >> DIO_SKIP_DIO_COUNT); >> if (ret < 0 && ret != -EIOCBQUEUED && ret_saved) >> return ret_saved; >> >> There's a lot of special casing here, so you might consider adding >> comments. > > Correct - maybe we should reconsider wrapper-izing this? :) Another option is just to skip dax_do_io() and this special casing fallback entirely if errors are present. I.e. only attempt dax_do_io when: IS_DAX() && gendisk->bb && bb->count == 0.
Dan Williams <dan.j.williams@intel.com> writes: >>> There's a lot of special casing here, so you might consider adding >>> comments. >> >> Correct - maybe we should reconsider wrapper-izing this? :) > > Another option is just to skip dax_do_io() and this special casing > fallback entirely if errors are present. I.e. only attempt dax_do_io > when: IS_DAX() && gendisk->bb && bb->count == 0. So, if there's an error anywhere on the device, penalize all I/O (not just writes, and not just on sectors that are bad)? I'm not sure that's a great plan, either. -Jeff
On Fri, Apr 15, 2016 at 11:06 AM, Jeff Moyer <jmoyer@redhat.com> wrote: > Dan Williams <dan.j.williams@intel.com> writes: > >>>> There's a lot of special casing here, so you might consider adding >>>> comments. >>> >>> Correct - maybe we should reconsider wrapper-izing this? :) >> >> Another option is just to skip dax_do_io() and this special casing >> fallback entirely if errors are present. I.e. only attempt dax_do_io >> when: IS_DAX() && gendisk->bb && bb->count == 0. > > So, if there's an error anywhere on the device, penalize all I/O (not > just writes, and not just on sectors that are bad)? I'm not sure that's > a great plan, either. > If errors are rare how much are we actually losing in practice? Moreover, we're going to do the full badblocks lookup anyway when we call ->direct_access(). If we had that information earlier we can avoid this fallback dance.
Dan Williams <dan.j.williams@intel.com> writes: > On Fri, Apr 15, 2016 at 11:06 AM, Jeff Moyer <jmoyer@redhat.com> wrote: >> Dan Williams <dan.j.williams@intel.com> writes: >> >>>>> There's a lot of special casing here, so you might consider adding >>>>> comments. >>>> >>>> Correct - maybe we should reconsider wrapper-izing this? :) >>> >>> Another option is just to skip dax_do_io() and this special casing >>> fallback entirely if errors are present. I.e. only attempt dax_do_io >>> when: IS_DAX() && gendisk->bb && bb->count == 0. >> >> So, if there's an error anywhere on the device, penalize all I/O (not >> just writes, and not just on sectors that are bad)? I'm not sure that's >> a great plan, either. >> > > If errors are rare how much are we actually losing in practice? How long is a piece of string? > Moreover, we're going to do the full badblocks lookup anyway when we > call ->direct_access(). If we had that information earlier we can > avoid this fallback dance. None of the proposed approaches looks clean to me. I'll go along with whatever you guys think is best. I am in favor of wrapping up all that duplicated code, though. Cheers, Jeff
On Fri, Apr 15, 2016 at 11:24 AM, Jeff Moyer <jmoyer@redhat.com> wrote: >> Moreover, we're going to do the full badblocks lookup anyway when we >> call ->direct_access(). If we had that information earlier we can >> avoid this fallback dance. > > None of the proposed approaches looks clean to me. I'll go along with > whatever you guys think is best. I am in favor of wrapping up all that > duplicated code, though. Christoph originally pushed for open coding this fallback decision per-filesystem. I agree with you on the "none the above" options are clean.
On Fri, 2016-04-15 at 11:17 -0700, Dan Williams wrote: > On Fri, Apr 15, 2016 at 11:06 AM, Jeff Moyer <jmoyer@redhat.com> wrote: > > > > Dan Williams <dan.j.williams@intel.com> writes: > > > > > > > There's a lot of special casing here, so you might consider > > > > > adding comments. > > > > Correct - maybe we should reconsider wrapper-izing this? :) > > > Another option is just to skip dax_do_io() and this special casing > > > fallback entirely if errors are present. I.e. only attempt dax_do_io > > > when: IS_DAX() && gendisk->bb && bb->count == 0. > > > > So, if there's an error anywhere on the device, penalize all I/O (not > > just writes, and not just on sectors that are bad)? I'm not sure > > that's a great plan, either. > > > If errors are rare how much are we actually losing in practice? > Moreover, we're going to do the full badblocks lookup anyway when we > call ->direct_access(). If we had that information earlier we can > avoid this fallback dance. A system running with DAX may have active data set in NVDIMM lager than RAM size. In this case, falling back to non-DAX will allocate page cache for the data, which will saturate the system with memory pressure. Thanks, -Toshi
On Fri, 2016-04-15 at 13:01 -0600, Toshi Kani wrote: > On Fri, 2016-04-15 at 11:17 -0700, Dan Williams wrote: > > > > On Fri, Apr 15, 2016 at 11:06 AM, Jeff Moyer <jmoyer@redhat.com> wrote: > > > > > > Dan Williams <dan.j.williams@intel.com> writes: > > > > > > > > > There's a lot of special casing here, so you might consider > > > > > > adding comments. > > > > > Correct - maybe we should reconsider wrapper-izing this? :) > > > > Another option is just to skip dax_do_io() and this special casing > > > > fallback entirely if errors are present. I.e. only attempt > > > > dax_do_io when: IS_DAX() && gendisk->bb && bb->count == 0. > > > > > > So, if there's an error anywhere on the device, penalize all I/O (not > > > just writes, and not just on sectors that are bad)? I'm not sure > > > that's a great plan, either. > > > > > If errors are rare how much are we actually losing in practice? > > Moreover, we're going to do the full badblocks lookup anyway when we > > call ->direct_access(). If we had that information earlier we can > > avoid this fallback dance. > > A system running with DAX may have active data set in NVDIMM lager than > RAM size. In this case, falling back to non-DAX will allocate page cache > for the data, which will saturate the system with memory pressure. Oh, sorry, we are still in DIO path. Falling back to DIO should not cause this issue. -Toshi
Dan Williams <dan.j.williams@intel.com> writes: > On Fri, Apr 15, 2016 at 11:24 AM, Jeff Moyer <jmoyer@redhat.com> wrote: >>> Moreover, we're going to do the full badblocks lookup anyway when we >>> call ->direct_access(). If we had that information earlier we can >>> avoid this fallback dance. >> >> None of the proposed approaches looks clean to me. I'll go along with >> whatever you guys think is best. I am in favor of wrapping up all that >> duplicated code, though. > > Christoph originally pushed for open coding this fallback decision > per-filesystem. I agree with you on the "none the above" options are > clean. I don't recall him saying "open code". Rather, the sentiment was to leave the fallback to the callers. That doesn't mean you can't wrap it up in a convenience function. Cheers, Jeff
On Fri, Apr 15, 2016 at 12:11:36PM -0400, Jeff Moyer wrote: > > + if (IS_DAX(inode)) { > > + ret = dax_do_io(iocb, inode, iter, offset, blkdev_get_block, > > NULL, DIO_SKIP_DIO_COUNT); > > + if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) > > + ret_saved = ret; > > + else > > + return ret; > > + } > > + > > + ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, offset, > > blkdev_get_block, NULL, NULL, > > DIO_SKIP_DIO_COUNT); > > + if (ret < 0 && ret_saved) > > + return ret_saved; > > + > > Hmm, did you just break async DIO? I think you did! :) > __blockdev_direct_IO can return -EIOCBQUEUED, and you've now turned that > into -EIO. Really, I don't see a reason to save that first -EIO. The > same applies to all instances in this patch. Yes, there is no point in saving the earlier error - just return the second error all the time. E.g. ret = dax_io(); if (dax_need_dio_retry(ret)) ret = direct_IO();
On Wed, 2016-04-20 at 13:59 -0700, Christoph Hellwig wrote: > On Fri, Apr 15, 2016 at 12:11:36PM -0400, Jeff Moyer wrote: > > > > > > > > + if (IS_DAX(inode)) { > > > + ret = dax_do_io(iocb, inode, iter, offset, > > > blkdev_get_block, > > > NULL, DIO_SKIP_DIO_COUNT); > > > + if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) > > > + ret_saved = ret; > > > + else > > > + return ret; > > > + } > > > + > > > + ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), > > > iter, offset, > > > blkdev_get_block, NULL, > > > NULL, > > > DIO_SKIP_DIO_COUNT); > > > + if (ret < 0 && ret_saved) > > > + return ret_saved; > > > + > > Hmm, did you just break async DIO? I think you did! :) > > __blockdev_direct_IO can return -EIOCBQUEUED, and you've now turned > > that > > into -EIO. Really, I don't see a reason to save that first > > -EIO. The > > same applies to all instances in this patch. > Yes, there is no point in saving the earlier error - just return the > second error all the time. Is it ok to do that? direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM due to some allocation failing, and I thought we should return the original -EIO in such cases so that the application doesn't lose the information that the bad block is actually causing the error. > > E.g. > > ret = dax_io(); > if (dax_need_dio_retry(ret)) > ret = direct_IO(); >
On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: > direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM due > to some allocation failing, and I thought we should return the original > -EIO in such cases so that the application doesn't lose the information > that the bad block is actually causing the error. EINVAL is a concern here. Not due to the right error reported, but because it means your current scheme is fundamentally broken - we need to support I/O at any alignment for DAX I/O, and not fail due to alignbment concernes for a highly specific degraded case. I think this whole series need to go back to the drawing board as I don't think it can actually rely on using direct I/O as the EIO fallback.
"hch@infradead.org" <hch@infradead.org> writes: > On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: >> direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM due >> to some allocation failing, and I thought we should return the original >> -EIO in such cases so that the application doesn't lose the information >> that the bad block is actually causing the error. > > EINVAL is a concern here. Not due to the right error reported, but > because it means your current scheme is fundamentally broken - we > need to support I/O at any alignment for DAX I/O, and not fail due to > alignbment concernes for a highly specific degraded case. > > I think this whole series need to go back to the drawing board as I > don't think it can actually rely on using direct I/O as the EIO > fallback. The only callers of dax_do_io are direct_IO methods. Cheers, Jeff
On Mon, 2016-04-25 at 01:31 -0700, hch@infradead.org wrote: > On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: > > > > direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM > > due > > to some allocation failing, and I thought we should return the > > original > > -EIO in such cases so that the application doesn't lose the > > information > > that the bad block is actually causing the error. > EINVAL is a concern here. Not due to the right error reported, but > because it means your current scheme is fundamentally broken - we > need to support I/O at any alignment for DAX I/O, and not fail due to > alignbment concernes for a highly specific degraded case. > > I think this whole series need to go back to the drawing board as I > don't think it can actually rely on using direct I/O as the EIO > fallback. > Agreed that DAX I/O can happen with any size/alignment, but how else do we send an IO through the driver without alignment restrictions? Also, the granularity at which we store badblocks is 512B sectors, so it seems natural that to clear such a sector, you'd expect to send a write to the whole sector. The expected usage flow is: - Application hits EIO doing dax_IO or load/store io - It checks badblocks and discovers it's files have lost data - It write()s those sectors (possibly converted to file offsets using fiemap) * This triggers the fallback path, but if the application is doing this level of recovery, it will know the sector is bad, and write the entire sector - Or it replaces the entire file from backup also using write() (not mmap+stores) * This just frees the fs block, and the next time the block is reallocated by the fs, it will likely be zeroed first, and that will be done through the driver and will clear errors I think if we want to keep allowing arbitrary alignments for the dax_do_io path, we'd need: 1. To represent badblocks at a finer granularity (likely cache lines) 2. To allow the driver to do IO to a *block device* at sub-sector granularity Can we do that?
On Mon, Apr 25, 2016 at 10:14 AM, Verma, Vishal L <vishal.l.verma@intel.com> wrote: > On Mon, 2016-04-25 at 01:31 -0700, hch@infradead.org wrote: >> On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: >> > >> > direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM >> > due >> > to some allocation failing, and I thought we should return the >> > original >> > -EIO in such cases so that the application doesn't lose the >> > information >> > that the bad block is actually causing the error. >> EINVAL is a concern here. Not due to the right error reported, but >> because it means your current scheme is fundamentally broken - we >> need to support I/O at any alignment for DAX I/O, and not fail due to >> alignbment concernes for a highly specific degraded case. >> >> I think this whole series need to go back to the drawing board as I >> don't think it can actually rely on using direct I/O as the EIO >> fallback. >> > Agreed that DAX I/O can happen with any size/alignment, but how else do > we send an IO through the driver without alignment restrictions? Also, > the granularity at which we store badblocks is 512B sectors, so it > seems natural that to clear such a sector, you'd expect to send a write > to the whole sector. > > The expected usage flow is: > > - Application hits EIO doing dax_IO or load/store io > > - It checks badblocks and discovers it's files have lost data > > - It write()s those sectors (possibly converted to file offsets using > fiemap) > * This triggers the fallback path, but if the application is doing > this level of recovery, it will know the sector is bad, and write the > entire sector > > - Or it replaces the entire file from backup also using write() (not > mmap+stores) > * This just frees the fs block, and the next time the block is > reallocated by the fs, it will likely be zeroed first, and that will be > done through the driver and will clear errors > > > I think if we want to keep allowing arbitrary alignments for the > dax_do_io path, we'd need: > 1. To represent badblocks at a finer granularity (likely cache lines) > 2. To allow the driver to do IO to a *block device* at sub-sector > granularity 3. Arrange for O_DIRECT to bypass dax_do_io(), and leave the optimization only for the dax "buffered I/O" case. 4. Skip dax_do_io() entirely in the presence of errors I think 3 is the most closely aligned with the typical block device model. In the typical case a buffered write may fail due to a badblock read when filling the page cache, but an O_DIRECT write would bypass the page cache and potentially clear the error / cause the block to be reallocated internally to the drive.
On Mon, Apr 25, 2016 at 05:14:36PM +0000, Verma, Vishal L wrote: > On Mon, 2016-04-25 at 01:31 -0700, hch@infradead.org wrote: > > On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: > > > > > > direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM > > > due > > > to some allocation failing, and I thought we should return the > > > original > > > -EIO in such cases so that the application doesn't lose the > > > information > > > that the bad block is actually causing the error. > > EINVAL is a concern here. Not due to the right error reported, but > > because it means your current scheme is fundamentally broken - we > > need to support I/O at any alignment for DAX I/O, and not fail due to > > alignbment concernes for a highly specific degraded case. > > > > I think this whole series need to go back to the drawing board as I > > don't think it can actually rely on using direct I/O as the EIO > > fallback. > > > Agreed that DAX I/O can happen with any size/alignment, but how else do > we send an IO through the driver without alignment restrictions? Also, > the granularity at which we store badblocks is 512B sectors, so it > seems natural that to clear such a sector, you'd expect to send a write > to the whole sector. > > The expected usage flow is: > > - Application hits EIO doing dax_IO or load/store io > > - It checks badblocks and discovers it's files have lost data Lots of hand-waving here. How does the application map a bad "sector" to a file without scanning the entire filesystem to find the owner of the bad sector? > - It write()s those sectors (possibly converted to file offsets using > fiemap) > * This triggers the fallback path, but if the application is doing > this level of recovery, it will know the sector is bad, and write the > entire sector Where does the application find the data that was lost to be able to rewrite it? > - Or it replaces the entire file from backup also using write() (not > mmap+stores) > * This just frees the fs block, and the next time the block is > reallocated by the fs, it will likely be zeroed first, and that will be > done through the driver and will clear errors There's an implicit assumption that applications will keep redundant copies of their data at the /application layer/ and be able to automatically repair it? And then there's the implicit assumption that it will unlink and free the entire file before writing a new copy, and that then assumes the the filesystem will zero blocks if they get reused to clear errors on that LBA sector mapping before they are accessible again to userspace.. It seems to me that there are a number of assumptions being made across multiple layers here. Maybe I've missed something - can you point me to the design/architecture description so I can see how "app does data recovery itself" dance is supposed to work? Cheers, Dave.
On Tue, Apr 26, 2016 at 09:25:52AM +1000, Dave Chinner wrote: > On Mon, Apr 25, 2016 at 05:14:36PM +0000, Verma, Vishal L wrote: > > On Mon, 2016-04-25 at 01:31 -0700, hch@infradead.org wrote: > > > On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: > > > > > > > > direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM > > > > due > > > > to some allocation failing, and I thought we should return the > > > > original > > > > -EIO in such cases so that the application doesn't lose the > > > > information > > > > that the bad block is actually causing the error. > > > EINVAL is a concern here. Not due to the right error reported, but > > > because it means your current scheme is fundamentally broken - we > > > need to support I/O at any alignment for DAX I/O, and not fail due to > > > alignbment concernes for a highly specific degraded case. > > > > > > I think this whole series need to go back to the drawing board as I > > > don't think it can actually rely on using direct I/O as the EIO > > > fallback. > > > > > Agreed that DAX I/O can happen with any size/alignment, but how else do > > we send an IO through the driver without alignment restrictions? Also, > > the granularity at which we store badblocks is 512B sectors, so it > > seems natural that to clear such a sector, you'd expect to send a write > > to the whole sector. > > > > The expected usage flow is: > > > > - Application hits EIO doing dax_IO or load/store io > > > > - It checks badblocks and discovers it's files have lost data > > Lots of hand-waving here. How does the application map a bad > "sector" to a file without scanning the entire filesystem to find > the owner of the bad sector? FWIW there was some discussion @ LSF about using (XFS) rmap to figure out which parts of a file (on XFS) have gone bad. Chris Mason said that he'd like to collaborate on having a common getfsmap ioctl between btrfs and XFS since they have a backref index that could be hooked up to it for them. Obviously the app still has to coordinate stopping file IO and calling GETFSMAP since the fs won't do that on its own. There's also the question of how to handle LBA translation if there's other stuff like dm in the way. I don't think device-mapper or md do reverse mapping, so things get murky from here. Guess I should get on pushing out a getfsmap patch for review. :) --D (/me doesn't have answers to any of your other questions.) > > - It write()s those sectors (possibly converted to file offsets using > > fiemap) > > * This triggers the fallback path, but if the application is doing > > this level of recovery, it will know the sector is bad, and write the > > entire sector > > Where does the application find the data that was lost to be able to > rewrite it? > > > - Or it replaces the entire file from backup also using write() (not > > mmap+stores) > > * This just frees the fs block, and the next time the block is > > reallocated by the fs, it will likely be zeroed first, and that will be > > done through the driver and will clear errors > > There's an implicit assumption that applications will keep redundant > copies of their data at the /application layer/ and be able to > automatically repair it? And then there's the implicit assumption > that it will unlink and free the entire file before writing a new > copy, and that then assumes the the filesystem will zero blocks if > they get reused to clear errors on that LBA sector mapping before > they are accessible again to userspace.. > > It seems to me that there are a number of assumptions being made > across multiple layers here. Maybe I've missed something - can you > point me to the design/architecture description so I can see how > "app does data recovery itself" dance is supposed to work? > > Cheers, > > Dave. > -- > Dave Chinner > david@fromorbit.com > > _______________________________________________ > xfs mailing list > xfs@oss.sgi.com > http://oss.sgi.com/mailman/listinfo/xfs
On Mon, Apr 25, 2016 at 4:25 PM, Dave Chinner <david@fromorbit.com> wrote: > On Mon, Apr 25, 2016 at 05:14:36PM +0000, Verma, Vishal L wrote: >> On Mon, 2016-04-25 at 01:31 -0700, hch@infradead.org wrote: >> > On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: >> > > >> > > direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM >> > > due >> > > to some allocation failing, and I thought we should return the >> > > original >> > > -EIO in such cases so that the application doesn't lose the >> > > information >> > > that the bad block is actually causing the error. >> > EINVAL is a concern here. Not due to the right error reported, but >> > because it means your current scheme is fundamentally broken - we >> > need to support I/O at any alignment for DAX I/O, and not fail due to >> > alignbment concernes for a highly specific degraded case. >> > >> > I think this whole series need to go back to the drawing board as I >> > don't think it can actually rely on using direct I/O as the EIO >> > fallback. >> > >> Agreed that DAX I/O can happen with any size/alignment, but how else do >> we send an IO through the driver without alignment restrictions? Also, >> the granularity at which we store badblocks is 512B sectors, so it >> seems natural that to clear such a sector, you'd expect to send a write >> to the whole sector. >> >> The expected usage flow is: >> >> - Application hits EIO doing dax_IO or load/store io >> >> - It checks badblocks and discovers it's files have lost data > > Lots of hand-waving here. How does the application map a bad > "sector" to a file without scanning the entire filesystem to find > the owner of the bad sector? > >> - It write()s those sectors (possibly converted to file offsets using >> fiemap) >> * This triggers the fallback path, but if the application is doing >> this level of recovery, it will know the sector is bad, and write the >> entire sector > > Where does the application find the data that was lost to be able to > rewrite it? > >> - Or it replaces the entire file from backup also using write() (not >> mmap+stores) >> * This just frees the fs block, and the next time the block is >> reallocated by the fs, it will likely be zeroed first, and that will be >> done through the driver and will clear errors > > There's an implicit assumption that applications will keep redundant > copies of their data at the /application layer/ and be able to > automatically repair it? And then there's the implicit assumption > that it will unlink and free the entire file before writing a new > copy, and that then assumes the the filesystem will zero blocks if > they get reused to clear errors on that LBA sector mapping before > they are accessible again to userspace.. > > It seems to me that there are a number of assumptions being made > across multiple layers here. Maybe I've missed something - can you > point me to the design/architecture description so I can see how > "app does data recovery itself" dance is supposed to work? > Maybe I missed something, but all these assumptions are already present for typical block devices, i.e. sectors may go bad and a write may make the sector usable again. This patch series is extending that out to the DAX-mmap case, but it's the same principle of "write to clear error" that we live with in the block-I/O path. What clarification are you looking for beyond that point?
On Tue, 2016-04-26 at 09:25 +1000, Dave Chinner wrote: > <> > > > > - It checks badblocks and discovers it's files have lost data > Lots of hand-waving here. How does the application map a bad > "sector" to a file without scanning the entire filesystem to find > the owner of the bad sector? Yes this was hand-wavey, but we talked about this a bit at LSF.. The idea is that a per-block-device badblocks list is available at /sys/block/<pmemX>/badblocks. The application (or a suitable yet-to-be- written library function) does a fiemap to figure out the sectors its files are using, and correlates the two lists. We can also look into providing an easier-to-use interface from the kernel, in the form of an fiemap flag to report only the bad sectors, or a SEEK_BAD flag.. The application doesn't have to scan the entire filesystem, but presumably it knows what files it 'owns', and does a fiemap for those. > > > > > - It write()s those sectors (possibly converted to file offsets > > using > > fiemap) > > * This triggers the fallback path, but if the application is > > doing > > this level of recovery, it will know the sector is bad, and write > > the > > entire sector > Where does the application find the data that was lost to be able to > rewrite it? The data that was lost is gone -- this assumes the application has some ability to recover using a journal/log or other redundancy - yes, at the application layer. If it doesn't have this sort of capability, the only option is to restore files from a backup/mirror. > > > > > - Or it replaces the entire file from backup also using write() (not > > mmap+stores) > > * This just frees the fs block, and the next time the block is > > reallocated by the fs, it will likely be zeroed first, and that will > > be > > done through the driver and will clear errors > There's an implicit assumption that applications will keep redundant > copies of their data at the /application layer/ and be able to > automatically repair it? And then there's the implicit assumption > that it will unlink and free the entire file before writing a new > copy, and that then assumes the the filesystem will zero blocks if > they get reused to clear errors on that LBA sector mapping before > they are accessible again to userspace.. > > It seems to me that there are a number of assumptions being made > across multiple layers here. Maybe I've missed something - can you > point me to the design/architecture description so I can see how > "app does data recovery itself" dance is supposed to work? There isn't a document other than the flow in my head :) - but maybe I could write one up.. I wasn't thinking the application itself maintains and restores from backup copy of the file.. The application hits either a SIGBUS or EIO depending on how it accesses the data, and crashes or raises some alarm. The recovery is then done out-of-band, by a sysadmin or such (i.e. delete the file, replace with a known good copy, restart application). To summarize, the two cases we want to handle are: 1. Application has inbuilt recovery: - hits badblock - figures out it is able to recover the data - handles SIGBUS or EIO - does a (sector aligned) write() to restore the data 2. Application doesn't have any inbuilt recovery mechanism - hits badblock - gets SIGBUS (or EIO) and crashes - Sysadmin restores file from backup Case 1 is handled by either a fallback to direct_IO from dax_do_io, or always _actually_ doing direct_IO when we're opened with O_DIRECT in spite of dax (what Dan suggested). Currently if we're mounted with dax, all IO O_DIRECT or otherwise will go through dax_do_io. Case 2 is handled by patch 4 of the series: dax: use sb_issue_zerout instead of calling dax_clear_sectors > > Cheers, > > Dave.
On Mon, Apr 25, 2016 at 04:43:14PM -0700, Dan Williams wrote: > On Mon, Apr 25, 2016 at 4:25 PM, Dave Chinner <david@fromorbit.com> wrote: > > On Mon, Apr 25, 2016 at 05:14:36PM +0000, Verma, Vishal L wrote: > >> On Mon, 2016-04-25 at 01:31 -0700, hch@infradead.org wrote: > >> > On Sat, Apr 23, 2016 at 06:08:37PM +0000, Verma, Vishal L wrote: > >> > > > >> > > direct_IO might fail with -EINVAL due to misalignment, or -ENOMEM > >> > > due > >> > > to some allocation failing, and I thought we should return the > >> > > original > >> > > -EIO in such cases so that the application doesn't lose the > >> > > information > >> > > that the bad block is actually causing the error. > >> > EINVAL is a concern here. Not due to the right error reported, but > >> > because it means your current scheme is fundamentally broken - we > >> > need to support I/O at any alignment for DAX I/O, and not fail due to > >> > alignbment concernes for a highly specific degraded case. > >> > > >> > I think this whole series need to go back to the drawing board as I > >> > don't think it can actually rely on using direct I/O as the EIO > >> > fallback. > >> > > >> Agreed that DAX I/O can happen with any size/alignment, but how else do > >> we send an IO through the driver without alignment restrictions? Also, > >> the granularity at which we store badblocks is 512B sectors, so it > >> seems natural that to clear such a sector, you'd expect to send a write > >> to the whole sector. > >> > >> The expected usage flow is: > >> > >> - Application hits EIO doing dax_IO or load/store io > >> > >> - It checks badblocks and discovers it's files have lost data > > > > Lots of hand-waving here. How does the application map a bad > > "sector" to a file without scanning the entire filesystem to find > > the owner of the bad sector? > > > >> - It write()s those sectors (possibly converted to file offsets using > >> fiemap) > >> * This triggers the fallback path, but if the application is doing > >> this level of recovery, it will know the sector is bad, and write the > >> entire sector > > > > Where does the application find the data that was lost to be able to > > rewrite it? > > > >> - Or it replaces the entire file from backup also using write() (not > >> mmap+stores) > >> * This just frees the fs block, and the next time the block is > >> reallocated by the fs, it will likely be zeroed first, and that will be > >> done through the driver and will clear errors > > > > There's an implicit assumption that applications will keep redundant > > copies of their data at the /application layer/ and be able to > > automatically repair it? And then there's the implicit assumption > > that it will unlink and free the entire file before writing a new > > copy, and that then assumes the the filesystem will zero blocks if > > they get reused to clear errors on that LBA sector mapping before > > they are accessible again to userspace.. > > > > It seems to me that there are a number of assumptions being made > > across multiple layers here. Maybe I've missed something - can you > > point me to the design/architecture description so I can see how > > "app does data recovery itself" dance is supposed to work? > > > > Maybe I missed something, but all these assumptions are already > present for typical block devices, i.e. sectors may go bad and a write > may make the sector usable again. The assumption we make about sectors going bad on SSDs or SRDs is that the device is about to die and needs replacing ASAP. Then RAID takes care of the rebuild completely transparently. i.e. handling and correcting bad sectors is typically done completely transparently /below/ the filesytem like so: Application Filesystem block [LBA mapping/redundancy/correction driver e.g. md/dm] driver hardware [LBA redundancy/correction e.g h/w RAID] In the case of filesystems with their own RAID/redundancy code (e.g. btrfs), then it looks like this: Application Filesystem mapping/redundancy/correction driver block driver hardware [LBA redundancy/correction e.g h/w RAID] > This patch series is extending that > out to the DAX-mmap case, but it's the same principle of "write to > clear error" that we live with in the block-I/O path. What > clarification are you looking for beyond that point? I'm asking for an actual design document that explains how moving all the redundancy and bad sector correction stuff from the LBA layer up into application space is supposed to work when applications have no clue about LBA mappings, nor tend to keep redundant data around. i.e. you're proposing this: Application Application data redundancy/correction Filesystem Block [LBA mapping/redundancy/correction driver e.g. md/dm] driver hardware And somehow all the error information from the hardware layer needs to be propagated up to the application layer, along with all the mapping information from the filesystem and block layers for the application to make sense of the hardware reported errors. I see assumptions this this "just works" but we don't have any of the relevant APIs or infrastructure to enable the application to do the hardware error->file+offset namespace mapping (i.e. filesystem reverse mapping for for file offsets and directory paths, and reverse mapping for the the block layer remapping drivers). I haven't seen any design/documentation for infrastructure at the application layer to handle redundant data and correctly transparently so I don't have any idea what the technical requirements this different IO stack places on filesystems may be. Hence I'm asking for some kind of architecture/design documentation that I can read to understand exactly what is being proposed here... Cheers, Dave.
On Mon, Apr 25, 2016 at 11:53:13PM +0000, Verma, Vishal L wrote: > On Tue, 2016-04-26 at 09:25 +1000, Dave Chinner wrote: > > > <> > > > > > > > - It checks badblocks and discovers it's files have lost data > > Lots of hand-waving here. How does the application map a bad > > "sector" to a file without scanning the entire filesystem to find > > the owner of the bad sector? > > Yes this was hand-wavey, but we talked about this a bit at LSF.. > The idea is that a per-block-device badblocks list is available at > /sys/block/<pmemX>/badblocks. The application (or a suitable yet-to-be- > written library function) does a fiemap to figure out the sectors its > files are using, and correlates the two lists. > We can also look into providing an easier-to-use interface from the > kernel, in the form of an fiemap flag to report only the bad sectors, or > a SEEK_BAD flag.. > The application doesn't have to scan the entire filesystem, but > presumably it knows what files it 'owns', and does a fiemap for those. You're assuming that only the DAX aware application accesses it's files. users, backup programs, data replicators, fileystem re-organisers (e.g. defragmenters) etc all may access the files and they may throw errors. What then? > > > - It write()s those sectors (possibly converted to file offsets > > > using > > > fiemap) > > > * This triggers the fallback path, but if the application is > > > doing > > > this level of recovery, it will know the sector is bad, and write > > > the > > > entire sector > > Where does the application find the data that was lost to be able to > > rewrite it? > > The data that was lost is gone -- this assumes the application has some > ability to recover using a journal/log or other redundancy - yes, at the > application layer. If it doesn't have this sort of capability, the only > option is to restore files from a backup/mirror. So the architecture has a built in assumption that only userspace can handle data loss? What about filesytsems like NOVA, that use log structured design to provide DAX w/ update atomicity and can potentially also provide redundancy/repair through the same mechanisms? Won't pmem native filesystems with built in data protection features like this remove the need for adding all this to userspace applications? If so, shouldn't that be the focus of development rahter than placing the burden on userspace apps to handle storage repair situations? > > > - Or it replaces the entire file from backup also using write() (not > > > mmap+stores) > > > * This just frees the fs block, and the next time the block is > > > reallocated by the fs, it will likely be zeroed first, and that will > > > be > > > done through the driver and will clear errors > > There's an implicit assumption that applications will keep redundant > > copies of their data at the /application layer/ and be able to > > automatically repair it? And then there's the implicit assumption > > that it will unlink and free the entire file before writing a new > > copy, and that then assumes the the filesystem will zero blocks if > > they get reused to clear errors on that LBA sector mapping before > > they are accessible again to userspace.. > > > > It seems to me that there are a number of assumptions being made > > across multiple layers here. Maybe I've missed something - can you > > point me to the design/architecture description so I can see how > > "app does data recovery itself" dance is supposed to work? > > There isn't a document other than the flow in my head :) - but maybe I > could write one up.. > I wasn't thinking the application itself maintains and restores from > backup copy of the file.. The application hits either a SIGBUS or EIO > depending on how it accesses the data, and crashes or raises some alarm. > The recovery is then done out-of-band, by a sysadmin or such (i.e. > delete the file, replace with a known good copy, restart application). > > To summarize, the two cases we want to handle are: > 1. Application has inbuilt recovery: > - hits badblock > - figures out it is able to recover the data > - handles SIGBUS or EIO > - does a (sector aligned) write() to restore the data The "figures out" step here is where >95% of the work we'd have to do is. And that's in filesystem and block layer code, not userspace, and userspace can't do that work in a signal handler. And it can still fall down to the second case when the application doesn't have another copy of the data somewhere. FWIW, we don't have a DAX enabled filesystem that can do reverse block mapping, so we're a year or two away from this being a workable production solution from the filesystem perspective. And AFAICT, it's not even on the roadmap for dm/md layers. > 2. Application doesn't have any inbuilt recovery mechanism > - hits badblock > - gets SIGBUS (or EIO) and crashes > - Sysadmin restores file from backup Which is no different to an existing non-DAX application getting an EIO/sigbus from current storage technologies. Except: in the existing storage stack, redundancy and correction has already had to have failed for the application to see such an error. Hence this is normally considered a DR case as there's had to be cascading failures (e.g. multiple disk failures in a RAID) to get to this stage, not a single error in a single sector in non-redundant storage. We need some form of redundancy and correction in the PMEM stack to prevent single sector errors from taking down services until an administrator can correct the problem. I'm trying to understand where this is supposed to fit into the picture - at this point I really don't think userspace applications are going to be able to do this reliably.... Cheers, Dave.
On Mon, Apr 25, 2016 at 5:11 PM, Dave Chinner <david@fromorbit.com> wrote: > On Mon, Apr 25, 2016 at 04:43:14PM -0700, Dan Williams wrote: [..] >> Maybe I missed something, but all these assumptions are already >> present for typical block devices, i.e. sectors may go bad and a write >> may make the sector usable again. > > The assumption we make about sectors going bad on SSDs or SRDs is > that the device is about to die and needs replacing ASAP. Similar assumptions here. Storage media is experiencing errors and past a certain threshold it may be time to decommission the device. You can see definitions for SMART / media health commands from various vendors at these links, and yes, hopefully these are standardized / unified at some point down the road: http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf https://github.com/HewlettPackard/hpe-nvm/blob/master/Documentation/NFIT_DSM_DDR4_NVDIMM-N_v84s.pdf https://msdn.microsoft.com/en-us/library/windows/hardware/mt604717(v=vs.85).aspx > Then > RAID takes care of the rebuild completely transparently. i.e. > handling and correcting bad sectors is typically done completely > transparently /below/ the filesytem like so: Again, same for an NVDIMM. Use the pmem block-device as a RAID-member device. >> This patch series is extending that >> out to the DAX-mmap case, but it's the same principle of "write to >> clear error" that we live with in the block-I/O path. What >> clarification are you looking for beyond that point? > > I'm asking for an actual design document that explains how moving > all the redundancy and bad sector correction stuff from the LBA > layer up into application space is supposed to work when > applications have no clue about LBA mappings, nor tend to keep > redundant data around. i.e. you're proposing this: These patches are not proposing *new* / general infrastructure for moving redundancy and bad sector correction handling to userspace. If an existing app is somehow dealing with raw (without RAID) device errors on disk storage media today it should not need to change to handle errors on an NVDIMM. My expectation is that very few if any applications handle this today and just fail in the presence of media errors. > Application > Application data redundancy/correction > Filesystem > Block > [LBA mapping/redundancy/correction driver e.g. md/dm] > driver > hardware > > And somehow all the error information from the hardware layer needs > to be propagated up to the application layer, along with all the > mapping information from the filesystem and block layers for the > application to make sense of the hardware reported errors. > > I see assumptions this this "just works" but we don't have any of > the relevant APIs or infrastructure to enable the application to do > the hardware error->file+offset namespace mapping (i.e. filesystem > reverse mapping for for file offsets and directory paths, and > reverse mapping for the the block layer remapping drivers). If an application expects errors to be handled beneath the filesystem then it should forgo DAX and arrange for the NVDIMM devices to be RAIDed. Otherwise, if an application wants to use DAX then it might need to be prepared to handle media errors itself same as the un-RAIDed disk case. Yes, at an administrative level without reverse-mapping support from a filesystem there's presently no way to ask "which files on this fs are impacted by media errors", and we're aware that reverse-mapping capabilities are nascent for current DAX-aware filesystems. The forward lookup path, as impractical as it is for large numbers of files, is available if an application wanted to know if a specific file was impacted. We've discussed possibly extending fiemap() to return bad blocks in a file rather than consulting sysfs, or extending lseek() with something like SEEK_ERROR to return offsets of bad areas in a file. > I haven't seen any design/documentation for infrastructure at the > application layer to handle redundant data and correctly > transparently so I don't have any idea what the technical > requirements this different IO stack places on filesystems may be. > Hence I'm asking for some kind of architecture/design documentation > that I can read to understand exactly what is being proposed here... I think this is a discussion for a solution that would build on top of this basic "here are the errors, re-write them with good data if you can; otherwise, best of luck" foundation. Something like a DAX-aware device mapper layer that duplicates data tagged with REQ_META so at least we have a recovery path when a sector error lands in critical filesystem-metadata. However, anything we come up with to make NVDIMM errors more survivable should be directly applicable to traditional disk storage as well. Along these lines we had a BoF session at Vault where drive vendors we're wondering if the sysfs bad sectors list could help software recover from the loss of a disk-head, or other errors that only take down part of the drive. An I/O hint that flags data that should be stored redundantly might be useful there as well. By the way, your presence was sorely missed at LSF/MM!
On Mon, Apr 25, 2016 at 06:45:08PM -0700, Dan Williams wrote: > On Mon, Apr 25, 2016 at 5:11 PM, Dave Chinner <david@fromorbit.com> wrote: > > On Mon, Apr 25, 2016 at 04:43:14PM -0700, Dan Williams wrote: > [..] > >> Maybe I missed something, but all these assumptions are already > >> present for typical block devices, i.e. sectors may go bad and a write > >> may make the sector usable again. > > > > The assumption we make about sectors going bad on SSDs or SRDs is > > that the device is about to die and needs replacing ASAP. > > Similar assumptions here. Storage media is experiencing errors and > past a certain threshold it may be time to decommission the device. > > You can see definitions for SMART / media health commands from various > vendors at these links, and yes, hopefully these are standardized / > unified at some point down the road: > > http://pmem.io/documents/NVDIMM_DSM_Interface_Example.pdf > https://github.com/HewlettPackard/hpe-nvm/blob/master/Documentation/NFIT_DSM_DDR4_NVDIMM-N_v84s.pdf > https://msdn.microsoft.com/en-us/library/windows/hardware/mt604717(v=vs.85).aspx > > > > Then > > RAID takes care of the rebuild completely transparently. i.e. > > handling and correcting bad sectors is typically done completely > > transparently /below/ the filesytem like so: > > Again, same for an NVDIMM. Use the pmem block-device as a RAID-member device. Which means we're not using DAX and so the existing storage model applies. I understand how this works. What I'm asking about the redundancy/error correction model /when using DAX/ and a userspace DAX load/store throws the MCE. > > And somehow all the error information from the hardware layer needs > > to be propagated up to the application layer, along with all the > > mapping information from the filesystem and block layers for the > > application to make sense of the hardware reported errors. > > > > I see assumptions this this "just works" but we don't have any of > > the relevant APIs or infrastructure to enable the application to do > > the hardware error->file+offset namespace mapping (i.e. filesystem > > reverse mapping for for file offsets and directory paths, and > > reverse mapping for the the block layer remapping drivers). > > If an application expects errors to be handled beneath the filesystem > then it should forgo DAX and arrange for the NVDIMM devices to be > RAIDed. See above: I'm asking about the DAX-enabled error handling model, not the traditional error handling model. > Otherwise, if an application wants to use DAX then it might > need to be prepared to handle media errors itself same as the > un-RAIDed disk case. Yes, at an administrative level without > reverse-mapping support from a filesystem there's presently no way to > ask "which files on this fs are impacted by media errors", and we're > aware that reverse-mapping capabilities are nascent for current > DAX-aware filesystems. Precisely my point - suggestions are being proposed which assume use of infrastructure that *does not exist yet* and has not been discussed or documented. If we're expecting such infrastructure to be implemented in the filesystems and block device drivers, then we need to determine that the error model actually works first... > The forward lookup path, as impractical as it > is for large numbers of files, is available if an application wanted > to know if a specific file was impacted. We've discussed possibly > extending fiemap() to return bad blocks in a file rather than > consulting sysfs, or extending lseek() with something like SEEK_ERROR > to return offsets of bad areas in a file. Via what infrastructure will the filesystem use for finding out whether a file has bad blocks in it? And if the file does have bad blocks, what are you expecting the filesystem to do with that information? > > I haven't seen any design/documentation for infrastructure at the > > application layer to handle redundant data and correctly > > transparently so I don't have any idea what the technical > > requirements this different IO stack places on filesystems may be. > > Hence I'm asking for some kind of architecture/design documentation > > that I can read to understand exactly what is being proposed here... > > I think this is a discussion for a solution that would build on top of > this basic "here are the errors, re-write them with good data if you > can; otherwise, best of luck" foundation. Something like a DAX-aware > device mapper layer that duplicates data tagged with REQ_META so at > least we have a recovery path when a sector error lands in critical > filesystem-metadata. Filesytsem metadata is not the topic of discussion here - it's user data that throws an error on a DAX load/store that is the issue. > However, anything we come up with to make NVDIMM > errors more survivable should be directly applicable to traditional > disk storage as well. I'm not sure it does. DAX implies that traditional block layer RAID infrastructure is not possible, nor are data CRCs, nor are any other sort of data transformations that are needed for redundancy at the device layers. Anything that relies on copying/modifying/stable data to provide redundancies needs to do such work at a place where it can stall userspace page faults. This is where pmem native filesystem designs like NOVA take over from traditional block based filesystems - they are designed around the ability to do atomic page-based operations for data protection and recovery operations. It is this mechanism that allows stable pages to be committed to permanent storage and as such, allow redundancy operations such as mirroring to be performed before operations are marked as "stable". I'm missing the bigger picture that is being aimed at here - what's the point of DAX if we have to turn it off if we want any sort of failure protection? What's the big plan for fully enabling DAX with robust error correction? Where is this all supposed to be leading to? > Along these lines we had a BoF session at Vault > where drive vendors we're wondering if the sysfs bad sectors list > could help software recover from the loss of a disk-head, or other > errors that only take down part of the drive. Right, but as I've said elsewhere, loss of a disk head implies terabyte scale data loss. That is not something we can automatically recovery from at the filesystem level. Low level raid recovery could handle that sort of loss, but at the higher layers it's a disaster similar to multiple disk RAID failure. It's a completely different scale to a single sector/page loss we are talking about here, and so I don't see there as being much (if any) overlap here. > An I/O hint that flags > data that should be stored redundantly might be useful there as well. DAX doesn't have an IO path to hint with... :/ Cheers, Dave.
On Mon, Apr 25, 2016 at 7:56 PM, Dave Chinner <david@fromorbit.com> wrote: > On Mon, Apr 25, 2016 at 06:45:08PM -0700, Dan Williams wrote: [..] >> Otherwise, if an application wants to use DAX then it might >> need to be prepared to handle media errors itself same as the >> un-RAIDed disk case. Yes, at an administrative level without >> reverse-mapping support from a filesystem there's presently no way to >> ask "which files on this fs are impacted by media errors", and we're >> aware that reverse-mapping capabilities are nascent for current >> DAX-aware filesystems. > > Precisely my point - suggestions are being proposed which assume > use of infrastructure that *does not exist yet* and has not been > discussed or documented. If we're expecting such infrastructure to > be implemented in the filesystems and block device drivers, then we > need to determine that the error model actually works first... These patches only assume the clear-error-on write-model, and that *maybe* the sysfs bad blocks list is useful if the filesystem has a reverse-map, or if the application can compare the list against the results of fiemap(). Beyond that, this is the same perennial "we should really have better error coordination between block device and filesystems" discussions that we have at LSF. > >> The forward lookup path, as impractical as it >> is for large numbers of files, is available if an application wanted >> to know if a specific file was impacted. We've discussed possibly >> extending fiemap() to return bad blocks in a file rather than >> consulting sysfs, or extending lseek() with something like SEEK_ERROR >> to return offsets of bad areas in a file. > > Via what infrastructure will the filesystem use for finding out > whether a file has bad blocks in it? And if the file does have bad > blocks, what are you expecting the filesystem to do with that > information? We currently have no expectation that the filesystem does anything with the bad blocks list. However, if a filesystem had btrfs-like capabilities to recover data from a redundant location we'd be looking to plug into that infrastructure. >> > I haven't seen any design/documentation for infrastructure at the >> > application layer to handle redundant data and correctly >> > transparently so I don't have any idea what the technical >> > requirements this different IO stack places on filesystems may be. >> > Hence I'm asking for some kind of architecture/design documentation >> > that I can read to understand exactly what is being proposed here... >> >> I think this is a discussion for a solution that would build on top of >> this basic "here are the errors, re-write them with good data if you >> can; otherwise, best of luck" foundation. Something like a DAX-aware >> device mapper layer that duplicates data tagged with REQ_META so at >> least we have a recovery path when a sector error lands in critical >> filesystem-metadata. > > Filesytsem metadata is not the topic of discussion here - it's > user data that throws an error on a DAX load/store that is the > issue. Which is not a new problem since volatile DRAM in the non-DAX case can throw the exact same error. The current recovery model there is crash the kernel (without MCE recovery), or crash the application and hope the kernel maps out the page or the application knows how to restart after SIGBUS. Memory mirroring is meant to make this a bit less harsh, but there's no mechanism to make this available outside the kernel. >> However, anything we come up with to make NVDIMM >> errors more survivable should be directly applicable to traditional >> disk storage as well. > > I'm not sure it does. DAX implies that traditional block layer RAID > infrastructure is not possible, nor are data CRCs, nor are any other > sort of data transformations that are needed for redundancy at the > device layers. Anything that relies on copying/modifying/stable data to > provide redundancies needs to do such work at a place where it can > stall userspace page faults. > > This is where pmem native filesystem designs like NOVA take over > from traditional block based filesystems - they are designed around > the ability to do atomic page-based operations for data protection > and recovery operations. It is this mechanism that allows stable > pages to be committed to permanent storage and as such, allow > redundancy operations such as mirroring to be performed before > operations are marked as "stable". > > I'm missing the bigger picture that is being aimed at here - what's the > point of DAX if we have to turn it off if we want any sort of > failure protection? What's the big plan for fully enabling DAX with > robust error correction? Where is this all supposed to be leading > to? > NOVA and other solutions are free and encouraged to do a coherent bottoms-up rethink of error handling on top of persistent memory devices, in the meantime applications can only expect the legacy SIGBUS and -EIO mechanisms are available. So I'm still trying to connect how the "What would NOVA do?" discussion is anything but orthogonal to hooking up SIGBUS and -EIO for traditional-filesystem DAX. It's the only error model an application can expect because it's the only one that currently exists. >> Along these lines we had a BoF session at Vault >> where drive vendors we're wondering if the sysfs bad sectors list >> could help software recover from the loss of a disk-head, or other >> errors that only take down part of the drive. > > Right, but as I've said elsewhere, loss of a disk head implies > terabyte scale data loss. That is not something we can automatically > recovery from at the filesystem level. Low level raid recovery could > handle that sort of loss, but at the higher layers it's a disaster > similar to multiple disk RAID failure. It's a completely different > scale to a single sector/page loss we are talking about here, and so > I don't see there as being much (if any) overlap here. > >> An I/O hint that flags >> data that should be stored redundantly might be useful there as well. > > DAX doesn't have an IO path to hint with... :/ ...I was thinking traditional filesystem metadata operations through the block layer. NOVA could of course do something better since it always indirects userspace access through a filesystem managed page.
On Mon, Apr 25, 2016 at 09:18:42PM -0700, Dan Williams wrote: > On Mon, Apr 25, 2016 at 7:56 PM, Dave Chinner <david@fromorbit.com> wrote: > > On Mon, Apr 25, 2016 at 06:45:08PM -0700, Dan Williams wrote: > >> > I haven't seen any design/documentation for infrastructure at the > >> > application layer to handle redundant data and correctly > >> > transparently so I don't have any idea what the technical > >> > requirements this different IO stack places on filesystems may be. > >> > Hence I'm asking for some kind of architecture/design documentation > >> > that I can read to understand exactly what is being proposed here... > >> > >> I think this is a discussion for a solution that would build on top of > >> this basic "here are the errors, re-write them with good data if you > >> can; otherwise, best of luck" foundation. Something like a DAX-aware > >> device mapper layer that duplicates data tagged with REQ_META so at > >> least we have a recovery path when a sector error lands in critical > >> filesystem-metadata. > > > > Filesytsem metadata is not the topic of discussion here - it's > > user data that throws an error on a DAX load/store that is the > > issue. > > Which is not a new problem since volatile DRAM in the non-DAX case can > throw the exact same error. They are not the same class of error, not by a long shot. The "bad page in page cache" error on traditional storage means data is not lost - the original copy still in whatever storage medium that the cached page was filled from. i.e. Re-read the file and the data is still there, which is no different to crashing and restarting that machine and losing whatever writes had not been committed to stable storage.. In the pmem case, a "bad page" is a permanent loss of data - it's unrecoverable without some form data recovery operation being performed on the storage. > The current recovery model there is crash > the kernel (without MCE recovery), Ouch. Permanent data loss and a system wide DoS. > or crash the application and hope > the kernel maps out the page or the application knows how to restart > after SIGBUS. Not much better - neither provide a mechanism for recovery. > Memory mirroring is meant to make this a bit less > harsh, but there's no mechanism to make this available outside the > kernel. Which implies that we need a DM module that interfaces with the hardware memory mirroring to perform recovery and remapping operations. i.e. in the traditional storage stack location. > >> However, anything we come up with to make NVDIMM > >> errors more survivable should be directly applicable to traditional > >> disk storage as well. > > > > I'm not sure it does. DAX implies that traditional block layer RAID > > infrastructure is not possible, nor are data CRCs, nor are any other > > sort of data transformations that are needed for redundancy at the > > device layers. Anything that relies on copying/modifying/stable data to > > provide redundancies needs to do such work at a place where it can > > stall userspace page faults. > > > > This is where pmem native filesystem designs like NOVA take over > > from traditional block based filesystems - they are designed around > > the ability to do atomic page-based operations for data protection > > and recovery operations. It is this mechanism that allows stable > > pages to be committed to permanent storage and as such, allow > > redundancy operations such as mirroring to be performed before > > operations are marked as "stable". > > > > I'm missing the bigger picture that is being aimed at here - what's the > > point of DAX if we have to turn it off if we want any sort of > > failure protection? What's the big plan for fully enabling DAX with > > robust error correction? Where is this all supposed to be leading > > to? > > > > NOVA and other solutions are free and encouraged to do a coherent > bottoms-up rethink of error handling on top of persistent memory > devices, in the meantime applications can only expect the legacy > SIGBUS and -EIO mechanisms are available. So I'm still trying to > connect how the "What would NOVA do?" discussion is anything but > orthogonal to hooking up SIGBUS and -EIO for traditional-filesystem > DAX. It's the only error model an application can expect because it's > the only one that currently exists. <sigh> Yes, I get that. I'm not interested in the resultant fatal error delivery - I'm asking about what happens between the memory error and the delivery of the fatal "we've lost your data forever" error that gets delivered to userspace. i.e. I'm after a description of how error correction/recovery is supposed to be applied to DAX *before we report SIGBUS or EIO* to the application. What is the plan/model/vision for intercepting MCEs and recovering from them? e.g. how do we going to pull the good copy from hardware/software memory mirrors? What layer is supposed to be responsible for that? Is it different for hardware mirroring compared to a more traditional software dm-RAID1 solution? What requirements does software recovery imply - do we need stable page state for DAX (i.e. to prevent userspace modification while we make copies)? Do we need to remap LBAs in the storage stack iduring recovery when bad blocks are reported? If so, where does it get done? What atomicity and resiliency requirements are there for recovery? e.g. bad block is reported, system crashes - what needs to happen on reboot to have recovery work correctly? There's heaps of stuff that is completely undefined here - error handling is fucking hard at the best of times, but I'm struggling to understand even the basics of what is being proposed here apart from "pmem error == crash the application, maybe even the system". Future filesystems are only part of the solution here - infrastructure like access to hardware mirrored copies for recovery purposes will impact greatly on the design of upper layers and their performance (e.g. no need for RAID1 in a software layer), so we really need the model/architecture to be pretty clearly defined at the outset before people waste too much time going down paths that simply won't work on the hardware/infrastructure that is being provided.... > >> An I/O hint that flags > >> data that should be stored redundantly might be useful there as well. > > > > DAX doesn't have an IO path to hint with... :/ > > ...I was thinking traditional filesystem metadata operations through > the block layer. NOVA could of course do something better since it > always indirects userspace access through a filesystem managed page. It seems to me you are focussing on code/technologies that exist today instead of trying to define an architecture that is more optimal for pmem storage systems. Yes, working code is great, but if you can't tell people how things like robust error handling and redundancy are going to work in future then it's going to take forever for everyone else to handle such errors robustly through the storage stack... Cheers, Dave.
On Mon, Apr 25, 2016 at 11:32:08AM -0400, Jeff Moyer wrote: > > EINVAL is a concern here. Not due to the right error reported, but > > because it means your current scheme is fundamentally broken - we > > need to support I/O at any alignment for DAX I/O, and not fail due to > > alignbment concernes for a highly specific degraded case. > > > > I think this whole series need to go back to the drawing board as I > > don't think it can actually rely on using direct I/O as the EIO > > fallback. > > The only callers of dax_do_io are direct_IO methods. They are because the DAX I/O pass is a mess, but that doesn't mean the user specific O_DIRECT on the open nessecarily.
On Mon, Apr 25, 2016 at 05:14:36PM +0000, Verma, Vishal L wrote: > - Application hits EIO doing dax_IO or load/store io > > - It checks badblocks and discovers it's files have lost data > > - It write()s those sectors (possibly converted to file offsets using > fiemap) > ?? ?? * This triggers the fallback path, but if the application is doing > this level of recovery, it will know the sector is bad, and write the > entire sector This sounds like a mess. > I think if we want to keep allowing arbitrary alignments for the > dax_do_io path, we'd need: > 1. To represent badblocks at a finer granularity (likely cache lines) > 2. To allow the driver to do IO to a *block device* at sub-sector > granularity It's not a block device if it supports DAX. It's byte addressable memory masquerading as a block device.
On Tue, 2016-04-26 at 10:41 +1000, Dave Chinner wrote: > <> > > The application doesn't have to scan the entire filesystem, but > > presumably it knows what files it 'owns', and does a fiemap for > > those. > You're assuming that only the DAX aware application accesses it's > files. users, backup programs, data replicators, fileystem > re-organisers (e.g. defragmenters) etc all may access the files and > they may throw errors. What then? In this scenario, backup applications etc that try to read that data before it has been replaced will just hit the errors and fail.. > <> > > The data that was lost is gone -- this assumes the application has > > some > > ability to recover using a journal/log or other redundancy - yes, > > at the > > application layer. If it doesn't have this sort of capability, the > > only > > option is to restore files from a backup/mirror. > So the architecture has a built in assumption that only userspace > can handle data loss? > > What about filesytsems like NOVA, that use log structured design to > provide DAX w/ update atomicity and can potentially also provide > redundancy/repair through the same mechanisms? Won't pmem native > filesystems with built in data protection features like this remove > the need for adding all this to userspace applications? > > If so, shouldn't that be the focus of development rahter than > placing the burden on userspace apps to handle storage repair > situations? Agreed that file systems like NOVA can be designed to handle this better, but haven't you said in the past that it may take years for a new file system to become production ready, and that DAX is the until- then solution that gets us most of the way there.. I think we just want to ensure that current-DAX has some way to deal with errors, and these patches provide an admin-intervention recovery path and possibly another if the app wants to try something fancy for recovery. <> > > > > > To summarize, the two cases we want to handle are: > > 1. Application has inbuilt recovery: > > - hits badblock > > - figures out it is able to recover the data > > - handles SIGBUS or EIO > > - does a (sector aligned) write() to restore the data > The "figures out" step here is where >95% of the work we'd have to > do is. And that's in filesystem and block layer code, not > userspace, and userspace can't do that work in a signal handler. > And it can still fall down to the second case when the application > doesn't have another copy of the data somewhere. Ah when I said "figures out" I was only thinking if the application has some redundancy/jouranlling, and if it can recover using that -- not additional recovery mechanisms at the block/fs layer. > > FWIW, we don't have a DAX enabled filesystem that can do > reverse block mapping, so we're a year or two away from this being a > workable production solution from the filesystem perspective. And > AFAICT, it's not even on the roadmap for dm/md layers. > > > > > 2. Application doesn't have any inbuilt recovery mechanism > > - hits badblock > > - gets SIGBUS (or EIO) and crashes > > - Sysadmin restores file from backup > Which is no different to an existing non-DAX application getting an > EIO/sigbus from current storage technologies. > > Except: in the existing storage stack, redundancy and correction has > already had to have failed for the application to see such an error. > Hence this is normally considered a DR case as there's had to be > cascading failures (e.g. multiple disk failures in a RAID) to get > to this stage, not a single error in a single sector in > non-redundant storage. > > We need some form of redundancy and correction in the PMEM stack to > prevent single sector errors from taking down services until an > administrator can correct the problem. I'm trying to understand > where this is supposed to fit into the picture - at this point I > really don't think userspace applications are going to be able to do > this reliably.... Agreed that the pmem stack could use more redundancy and error correction, perhaps enabling md-raid to raid pmem devices and then enable DAX on top of that and we'll have a better chance to handle errors, but that level of recovery isn't what these patches are aiming for -- that is obviously a longer term effort. These simply aim to provide that disaster recovery path when a single sector failure does take down the service. Today, on a dax enabled filesystem, if/when the app hits an error and crashes, dax is simply disabled till the errors are gone. This is obviously less than ideal. (This was done because there is currently no way for a DAX file system to send any IO - mmap or otherwise - through the driver, including zeroing of new fs blocks). These patches enable the DR path by allowing some non-mmap IO (most importantly zeroing) to go through the driver which can tell the device to do some remapping etc. So, yes, this is very much a DR case in our current pmem+dax architecture, and we should probably design more robust handling at the block/md/fs layer, but with these, you at least get to crash the app, delete its files and restore them from out-of-band backups and continue with DAX. > > Cheers, > > Dave.
On Tue, Apr 26, 2016 at 1:27 AM, Dave Chinner <david@fromorbit.com> wrote: > On Mon, Apr 25, 2016 at 09:18:42PM -0700, Dan Williams wrote: [..] > It seems to me you are focussing on code/technologies that exist > today instead of trying to define an architecture that is more > optimal for pmem storage systems. Yes, working code is great, but if > you can't tell people how things like robust error handling and > redundancy are going to work in future then it's going to take > forever for everyone else to handle such errors robustly through the > storage stack... Precisely because higher order redundancy is built on top this baseline. MD-RAID can't do it's error recovery if we don't have -EIO and clear-error-on-write. On the other hand, you're absolutely right that we have a gaping hole on top of the SIGBUS recovery model, and don't have a kernel layer we can interpose on top of DAX to provide some semblance of redundancy. In the meantime, a handful of applications with a team of full-time site-reliability-engineers may be able to plug in external redundancy infrastructure on top of what is defined in these patches. For everyone else, the hard problem, we need to do a lot more thinking about a trap and recover solution.
On Tue, 2016-04-26 at 01:33 -0700, hch@infradead.org wrote: > On Mon, Apr 25, 2016 at 05:14:36PM +0000, Verma, Vishal L wrote: > > > > - Application hits EIO doing dax_IO or load/store io > > > > - It checks badblocks and discovers it's files have lost data > > > > - It write()s those sectors (possibly converted to file offsets > > using > > fiemap) > > ?? ?? * This triggers the fallback path, but if the application is > > doing > > this level of recovery, it will know the sector is bad, and write > > the > > entire sector > This sounds like a mess. > > > > > I think if we want to keep allowing arbitrary alignments for the > > dax_do_io path, we'd need: > > 1. To represent badblocks at a finer granularity (likely cache > > lines) > > 2. To allow the driver to do IO to a *block device* at sub-sector > > granularity > It's not a block device if it supports DAX. It's byte addressable > memory masquerading as a block device. Yes but we made that decision a while back with pmem :) Are you saying it should stop being a block device anymore? > -- > To unsubscribe from this list: send the line "unsubscribe linux- > block" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html
On Tue 26-04-16 07:59:10, Dan Williams wrote: > On Tue, Apr 26, 2016 at 1:27 AM, Dave Chinner <david@fromorbit.com> wrote: > > On Mon, Apr 25, 2016 at 09:18:42PM -0700, Dan Williams wrote: > [..] > > It seems to me you are focussing on code/technologies that exist > > today instead of trying to define an architecture that is more > > optimal for pmem storage systems. Yes, working code is great, but if > > you can't tell people how things like robust error handling and > > redundancy are going to work in future then it's going to take > > forever for everyone else to handle such errors robustly through the > > storage stack... > > Precisely because higher order redundancy is built on top this baseline. > > MD-RAID can't do it's error recovery if we don't have -EIO and > clear-error-on-write. On the other hand, you're absolutely right that > we have a gaping hole on top of the SIGBUS recovery model, and don't > have a kernel layer we can interpose on top of DAX to provide some > semblance of redundancy. > > In the meantime, a handful of applications with a team of full-time > site-reliability-engineers may be able to plug in external redundancy > infrastructure on top of what is defined in these patches. For > everyone else, the hard problem, we need to do a lot more thinking > about a trap and recover solution. So we could actually implement some kind of redundancy with DAX with reasonable effort. We already do track dirty storage PFNs in the radix tree. After DAX locking patches get merged we also have a reliable way to write-protect them when we decide to do 'writeback' (translates to flushing CPU caches) for them. When we do that, we have all the infrastructure in place to provide 'stable pages' while some mirroring or other redundancy mechanism in kernel works with the data. But as Dave said, we should do some writeup of how this is all supposed to work and e.g. which layer is going to be responsible for the redundancy. Do we want to have that in DAX code? Or just provide stable page guarantees from DAX and do the redundancy from device mapper? This needs more thought... Honza
On Tue, Apr 26, 2016 at 8:31 AM, Jan Kara <jack@suse.cz> wrote: > On Tue 26-04-16 07:59:10, Dan Williams wrote: >> On Tue, Apr 26, 2016 at 1:27 AM, Dave Chinner <david@fromorbit.com> wrote: >> > On Mon, Apr 25, 2016 at 09:18:42PM -0700, Dan Williams wrote: >> [..] >> > It seems to me you are focussing on code/technologies that exist >> > today instead of trying to define an architecture that is more >> > optimal for pmem storage systems. Yes, working code is great, but if >> > you can't tell people how things like robust error handling and >> > redundancy are going to work in future then it's going to take >> > forever for everyone else to handle such errors robustly through the >> > storage stack... >> >> Precisely because higher order redundancy is built on top this baseline. >> >> MD-RAID can't do it's error recovery if we don't have -EIO and >> clear-error-on-write. On the other hand, you're absolutely right that >> we have a gaping hole on top of the SIGBUS recovery model, and don't >> have a kernel layer we can interpose on top of DAX to provide some >> semblance of redundancy. >> >> In the meantime, a handful of applications with a team of full-time >> site-reliability-engineers may be able to plug in external redundancy >> infrastructure on top of what is defined in these patches. For >> everyone else, the hard problem, we need to do a lot more thinking >> about a trap and recover solution. > > So we could actually implement some kind of redundancy with DAX with > reasonable effort. We already do track dirty storage PFNs in the radix > tree. After DAX locking patches get merged we also have a reliable way to > write-protect them when we decide to do 'writeback' (translates to flushing > CPU caches) for them. When we do that, we have all the infrastructure in > place to provide 'stable pages' while some mirroring or other redundancy > mechanism in kernel works with the data. > > But as Dave said, we should do some writeup of how this is all supposed to > work and e.g. which layer is going to be responsible for the redundancy. Do > we want to have that in DAX code? Or just provide stable page guarantees > from DAX and do the redundancy from device mapper? This needs more > thought... > [ adding Mike, since his ears are likely burning by this point ] If we had the ability to specify a range or list of ranges to blkdev_issue_flush() that would allow the driver level to implement redundancy at sync time. And no, before someone flies off the handle, this isn't rehashing the same argument I lost about where to track dirty pfns. Rather this relies on the radix to track dirty pfns, but asks the driver to do the flush operation. In the nominal case this is a clflush / clwb loop or wbinvd in the pmem driver, in the redundancy case the pmem driver is swapped out for a driver that uses the flush request as a trigger point to synchronize redundant data. We want this at the driver level to take advantage of standard asynchronous completions, and make it administratively equivalent to the dm/md layering people are used to using.
Dave Chinner <david@fromorbit.com> writes: > On Mon, Apr 25, 2016 at 11:53:13PM +0000, Verma, Vishal L wrote: >> On Tue, 2016-04-26 at 09:25 +1000, Dave Chinner wrote: > You're assuming that only the DAX aware application accesses it's > files. users, backup programs, data replicators, fileystem > re-organisers (e.g. defragmenters) etc all may access the files and > they may throw errors. What then? I'm not sure how this is any different from regular storage. If an application gets EIO, it's up to the app to decide what to do with that. >> > Where does the application find the data that was lost to be able to >> > rewrite it? >> >> The data that was lost is gone -- this assumes the application has some >> ability to recover using a journal/log or other redundancy - yes, at the >> application layer. If it doesn't have this sort of capability, the only >> option is to restore files from a backup/mirror. > > So the architecture has a built in assumption that only userspace > can handle data loss? Remember that the proposed programming model completely bypasses the kernel, so yes, it is expected that user-space will have to deal with the problem. > What about filesytsems like NOVA, that use log structured design to > provide DAX w/ update atomicity and can potentially also provide > redundancy/repair through the same mechanisms? Won't pmem native > filesystems with built in data protection features like this remove > the need for adding all this to userspace applications? I don't think we'll /only/ support NOVA for pmem. So we'll have to deal with this for existing file systems, right? > If so, shouldn't that be the focus of development rahter than > placing the burden on userspace apps to handle storage repair > situations? It really depends on the programming model. In the model Vishal is talking about, either the applications themselves or the libraries they link to are expected to implement the redundancies where necessary. >> > There's an implicit assumption that applications will keep redundant >> > copies of their data at the /application layer/ and be able to >> > automatically repair it? That's one way to do things. It really depends on the application what it will do for recovery. >> > And then there's the implicit assumption that it will unlink and >> > free the entire file before writing a new copy I think Vishal was referring to restoring from backup. cp itself will truncate the file before overwriting, iirc. >> To summarize, the two cases we want to handle are: >> 1. Application has inbuilt recovery: >> - hits badblock >> - figures out it is able to recover the data >> - handles SIGBUS or EIO >> - does a (sector aligned) write() to restore the data > > The "figures out" step here is where >95% of the work we'd have to > do is. And that's in filesystem and block layer code, not > userspace, and userspace can't do that work in a signal handler. > And it can still fall down to the second case when the application > doesn't have another copy of the data somewhere. I read that "figures out" step as the application determining whether or not it had a redundant copy. > FWIW, we don't have a DAX enabled filesystem that can do > reverse block mapping, so we're a year or two away from this being a > workable production solution from the filesystem perspective. And > AFAICT, it's not even on the roadmap for dm/md layers. Do we even need that? What if we added an FIEMAP flag for determining bad blocks. The file system could simply walk the list of extents for the file and check the corresponding disk blocks. No reverse mapping required. Also note that DM/MD don't support direct_access(), either, so I don't think they're relevant for this discussion. >> 2. Application doesn't have any inbuilt recovery mechanism >> - hits badblock >> - gets SIGBUS (or EIO) and crashes >> - Sysadmin restores file from backup > > Which is no different to an existing non-DAX application getting an > EIO/sigbus from current storage technologies. > > Except: in the existing storage stack, redundancy and correction has > already had to have failed for the application to see such an error. > Hence this is normally considered a DR case as there's had to be > cascading failures (e.g. multiple disk failures in a RAID) to get > to this stage, not a single error in a single sector in > non-redundant storage. > > We need some form of redundancy and correction in the PMEM stack to > prevent single sector errors from taking down services until an > administrator can correct the problem. I'm trying to understand > where this is supposed to fit into the picture - at this point I > really don't think userspace applications are going to be able to do > this reliably.... Not all storage is configured into a RAID volume, and in some instances, the application is better positioned to recover the data (gluster/ceph, for example). It really comes down to whether applications or libraries will want to implement redundancy themselves in order to get a bump in performance by not going through the kernel. And I think I know what your opinion is on that front. :-) Speaking of which, did you see the numbers Dan shared at LSF on how much overhead there is in calling into the kernel for syncing? Dan, can/did you publish that spreadsheet somewhere? Cheers, Jeff
On Mon, May 2, 2016 at 8:18 AM, Jeff Moyer <jmoyer@redhat.com> wrote: > Dave Chinner <david@fromorbit.com> writes: [..] >> We need some form of redundancy and correction in the PMEM stack to >> prevent single sector errors from taking down services until an >> administrator can correct the problem. I'm trying to understand >> where this is supposed to fit into the picture - at this point I >> really don't think userspace applications are going to be able to do >> this reliably.... > > Not all storage is configured into a RAID volume, and in some instances, > the application is better positioned to recover the data (gluster/ceph, > for example). It really comes down to whether applications or libraries > will want to implement redundancy themselves in order to get a bump in > performance by not going through the kernel. And I think I know what > your opinion is on that front. :-) > > Speaking of which, did you see the numbers Dan shared at LSF on how much > overhead there is in calling into the kernel for syncing? Dan, can/did > you publish that spreadsheet somewhere? Here it is: https://docs.google.com/spreadsheets/d/1pwr9psy6vtB9DOsc2bUdXevJRz5Guf6laZ4DaZlkhoo/edit?usp=sharing On the "Filtered" tab I have some of the comparisons where: noop => don't call msync and don't flush caches in userspace persist => cache flushing only in userspace and only on individual cache lines persist_4k => cache flushing only in userspace, but flushing is performed in 4K aligned units msync => same granularity flushing as the 'persist' case, but the kernel internally promotes this to a 4K sized / aligned flush msync_0 => synthetic case where msync() returns immediately and does no other work The takeaway is that msync() is 9-10x slower than userspace cache management. Let me know if there are any questions and I can add an NVML developer to this thread...
On Mon, May 02, 2016 at 11:18:36AM -0400, Jeff Moyer wrote: > Dave Chinner <david@fromorbit.com> writes: > > > On Mon, Apr 25, 2016 at 11:53:13PM +0000, Verma, Vishal L wrote: > >> On Tue, 2016-04-26 at 09:25 +1000, Dave Chinner wrote: > > You're assuming that only the DAX aware application accesses it's > > files. users, backup programs, data replicators, fileystem > > re-organisers (e.g. defragmenters) etc all may access the files and > > they may throw errors. What then? > > I'm not sure how this is any different from regular storage. If an > application gets EIO, it's up to the app to decide what to do with that. Sure - they'll fail. But the question I'm asking is that if the application that owns the data is supposed to do error recovery, what happens when a 3rd party application hits an error? If that consumes the error, the the app that owns the data won't ever get a chance to correct the error. This is a minefield - a 3rd party app that swallows and clears DAX based IO errors is a data corruption vector. can yo imagine if *grep* did this? The model that is being promoted here effectively allows this sort of behaviour - I don't really think we should be architecting an error recovery strategy that has the capability to go this wrong.... > >> > Where does the application find the data that was lost to be able to > >> > rewrite it? > >> > >> The data that was lost is gone -- this assumes the application has some > >> ability to recover using a journal/log or other redundancy - yes, at the > >> application layer. If it doesn't have this sort of capability, the only > >> option is to restore files from a backup/mirror. > > > > So the architecture has a built in assumption that only userspace > > can handle data loss? > > Remember that the proposed programming model completely bypasses the > kernel, so yes, it is expected that user-space will have to deal with > the problem. No, it doesn't completely bypass the kernel - the kernel is the infrastructure that catches the errors in the first place, and it owns and controls all the metadata that corresponds to the physical location of that error. The only thing the kernel doesn't own is the *contents* of that location. > > What about filesytsems like NOVA, that use log structured design to > > provide DAX w/ update atomicity and can potentially also provide > > redundancy/repair through the same mechanisms? Won't pmem native > > filesystems with built in data protection features like this remove > > the need for adding all this to userspace applications? > > I don't think we'll /only/ support NOVA for pmem. So we'll have to deal > with this for existing file systems, right? Yes, but that misses my point that it seems that the design is only focussed on userspace and existing filesystems and there is no consideration of kernel side functionality that could do transparent recovery.... > > If so, shouldn't that be the focus of development rahter than > > placing the burden on userspace apps to handle storage repair > > situations? > > It really depends on the programming model. In the model Vishal is > talking about, either the applications themselves or the libraries they > link to are expected to implement the redundancies where necessary. IOWs, filesystems no longer have any control over data integrity. Yet it's the filesystem developers who will still be responsible for data integrity and when the filesystem has a data coruption event we'll get blamed and the filesystem gets a bad name, even though it's entirely the applications fault. We've seen this time and time again - application developers cannot be trusted to guarantee data integrity. yes, some apps will be fine, but do you really expect application devs that refuse to use fsync because it's too slow are going to have a different approach to integrity when it comes to DAX? > >> > There's an implicit assumption that applications will keep redundant > >> > copies of their data at the /application layer/ and be able to > >> > automatically repair it? > > That's one way to do things. It really depends on the application what > it will do for recovery. > > >> > And then there's the implicit assumption that it will unlink and > >> > free the entire file before writing a new copy > > I think Vishal was referring to restoring from backup. cp itself will > truncate the file before overwriting, iirc. Which version of cp? what happens if they use --sparse and the error is in a zeroed region? There's so many assumptions about undefined userspace environment, application and user behaviour being made here, and it's all being handwaved away. I'm asking for this to be defined, demonstrated and documented as a working model that cannot be abused and doesn't have holes the size of trucks in it, not handwaving... > >> To summarize, the two cases we want to handle are: > >> 1. Application has inbuilt recovery: > >> - hits badblock > >> - figures out it is able to recover the data > >> - handles SIGBUS or EIO > >> - does a (sector aligned) write() to restore the data > > > > The "figures out" step here is where >95% of the work we'd have to > > do is. And that's in filesystem and block layer code, not > > userspace, and userspace can't do that work in a signal handler. > > And it can still fall down to the second case when the application > > doesn't have another copy of the data somewhere. > > I read that "figures out" step as the application determining whether or > not it had a redundant copy. Another undocumented assumption, that doesn't simplify what needs to be done. Indeed, userspace can't do that until it is in SIGBUS context, which tends to imply applications need to do a major amount of work from within the signal handler.... > > FWIW, we don't have a DAX enabled filesystem that can do > > reverse block mapping, so we're a year or two away from this being a > > workable production solution from the filesystem perspective. And > > AFAICT, it's not even on the roadmap for dm/md layers. > > Do we even need that? What if we added an FIEMAP flag for determining > bad blocks. So you're assuming that the filesystem has been informed of the bad blocks and has already marked the bad regions of the file in it's extent list? How does that happen? What mechanism is used for the underlying block device to inform the filesytem that theirs a bad LBA, and how does the filesytem the map that to a path/file/offset with reverse mapping? Or is there some other magic that hasn't been explained happening here? > The file system could simply walk the list of extents for > the file and check the corresponding disk blocks. No reverse mapping > required. You're expecting the filesystem to poll the block device to find bad sectors? Ignoring the fact this is the sort of brute force scan we need reverse mapping to avoid, how does the filesystem know what file/extent list it should be searching when the block device informs it there is a bad sector somewhere? i.e. what information does the MCE convey to the block device, and what does the block device pass to the filesytem so the filesystem can do one of these scans? If the block device is only passing LBAs or a generic "new bad block has been found" message, the filesystem still has to do an full scan of it's metadata to find the owner of the LBA(s) that have gone bad.... Nobody is explaining these important little details - there seems to be an assumption that everyone "knows" how this is all going to work and that we have infrastructure that can make it work. Just because we might be able to present bad block information to userspace via FIEMAP doesn't mean that it's trivial to implement. THE FIEMAP flag is trivial - connecting the dots is the hard part and nobody is explaining to me how that is all supposed to be done. > Also note that DM/MD don't support direct_access(), either, > so I don't think they're relevant for this discussion. But they could for linear concatenation, which would be extremely useful. e.g. stitching per-node non-linear pmem into a single linear LBA range.... > >> 2. Application doesn't have any inbuilt recovery mechanism > >> - hits badblock > >> - gets SIGBUS (or EIO) and crashes > >> - Sysadmin restores file from backup > > > > Which is no different to an existing non-DAX application getting an > > EIO/sigbus from current storage technologies. > > > > Except: in the existing storage stack, redundancy and correction has > > already had to have failed for the application to see such an error. > > Hence this is normally considered a DR case as there's had to be > > cascading failures (e.g. multiple disk failures in a RAID) to get > > to this stage, not a single error in a single sector in > > non-redundant storage. > > > > We need some form of redundancy and correction in the PMEM stack to > > prevent single sector errors from taking down services until an > > administrator can correct the problem. I'm trying to understand > > where this is supposed to fit into the picture - at this point I > > really don't think userspace applications are going to be able to do > > this reliably.... > > Not all storage is configured into a RAID volume, and in some instances, > the application is better positioned to recover the data (gluster/ceph, > for example). Right, but they still rely on the filesystem to provide data integrity guarantees to work correctly. While they have "node level" redundancy, operations within the node still need to work correctly and so they'd still need all the kernel/fs side functionality to provide them with error information (like fiemap bad blocks) on top of all the new error detectiona nd correction code they'd need to support this... FWIW, the whole point of DAX on existing filesystems was to not need major changes to existing filesystems to support fast pmem operations. i.e. to get something working quickly while pmem native filesytems are developed to support pmem and all it's quirks in a clean and efficient manner. Instead, what I'm seeing now is a trend towards forcing existing filesystems to support the requirements and quirks of DAX and pmem, without any focus on pmem native solutions. i.e. I'm hearing "we need major surgery to existing filesystems and block devices to make DAX work" rather than "how do we make this efficient for a pmem native solution rather than being bound to block device semantics"? Cheers, Dave.
On Tue, 2016-05-03 at 09:04 +1000, Dave Chinner wrote: > On Mon, May 02, 2016 at 11:18:36AM -0400, Jeff Moyer wrote: > > > > Dave Chinner <david@fromorbit.com> writes: > > > > > > > > On Mon, Apr 25, 2016 at 11:53:13PM +0000, Verma, Vishal L wrote: > > > > > > > > On Tue, 2016-04-26 at 09:25 +1000, Dave Chinner wrote: > > > You're assuming that only the DAX aware application accesses it's > > > files. users, backup programs, data replicators, fileystem > > > re-organisers (e.g. defragmenters) etc all may access the files > > > and > > > they may throw errors. What then? > > I'm not sure how this is any different from regular storage. If an > > application gets EIO, it's up to the app to decide what to do with > > that. > Sure - they'll fail. But the question I'm asking is that if the > application that owns the data is supposed to do error recovery, > what happens when a 3rd party application hits an error? If that > consumes the error, the the app that owns the data won't ever get a > chance to correct the error. > > This is a minefield - a 3rd party app that swallows and clears DAX > based IO errors is a data corruption vector. can yo imagine if > *grep* did this? The model that is being promoted here effectively > allows this sort of behaviour - I don't really think we > should be architecting an error recovery strategy that has the > capability to go this wrong.... > Just to address this bit - No. Any number of backup/3rd party application can hit the error and _fail_ but surely they won't try to _write_ the bad location? Only a write to the bad sector will clear it in this model - and until such time, all reads will just keep erroring out. This works for DAX/mmap based reads/writes too - mmap-stores won't/can't clear errors - you have to go through the block path, and in the altest version of my patch set, that has to be explicitly through O_DIRECT.
On Mon, May 2, 2016 at 4:04 PM, Dave Chinner <david@fromorbit.com> wrote: > On Mon, May 02, 2016 at 11:18:36AM -0400, Jeff Moyer wrote: >> Dave Chinner <david@fromorbit.com> writes: >> >> > On Mon, Apr 25, 2016 at 11:53:13PM +0000, Verma, Vishal L wrote: >> >> On Tue, 2016-04-26 at 09:25 +1000, Dave Chinner wrote: >> > You're assuming that only the DAX aware application accesses it's >> > files. users, backup programs, data replicators, fileystem >> > re-organisers (e.g. defragmenters) etc all may access the files and >> > they may throw errors. What then? >> >> I'm not sure how this is any different from regular storage. If an >> application gets EIO, it's up to the app to decide what to do with that. > > Sure - they'll fail. But the question I'm asking is that if the > application that owns the data is supposed to do error recovery, > what happens when a 3rd party application hits an error? If that > consumes the error, the the app that owns the data won't ever get a > chance to correct the error. > > This is a minefield - a 3rd party app that swallows and clears DAX > based IO errors is a data corruption vector. can yo imagine if > *grep* did this? The model that is being promoted here effectively > allows this sort of behaviour - I don't really think we > should be architecting an error recovery strategy that has the > capability to go this wrong.... Since when does grep write to a file on error? > >> >> > Where does the application find the data that was lost to be able to >> >> > rewrite it? >> >> >> >> The data that was lost is gone -- this assumes the application has some >> >> ability to recover using a journal/log or other redundancy - yes, at the >> >> application layer. If it doesn't have this sort of capability, the only >> >> option is to restore files from a backup/mirror. >> > >> > So the architecture has a built in assumption that only userspace >> > can handle data loss? >> >> Remember that the proposed programming model completely bypasses the >> kernel, so yes, it is expected that user-space will have to deal with >> the problem. > > No, it doesn't completely bypass the kernel - the kernel is the > infrastructure that catches the errors in the first place, and it > owns and controls all the metadata that corresponds to the physical > location of that error. The only thing the kernel doesn't own is the > *contents* of that location. > >> > What about filesytsems like NOVA, that use log structured design to >> > provide DAX w/ update atomicity and can potentially also provide >> > redundancy/repair through the same mechanisms? Won't pmem native >> > filesystems with built in data protection features like this remove >> > the need for adding all this to userspace applications? >> >> I don't think we'll /only/ support NOVA for pmem. So we'll have to deal >> with this for existing file systems, right? > > Yes, but that misses my point that it seems that the design is only > focussed on userspace and existing filesystems and there is no > consideration of kernel side functionality that could do transparent > recovery.... > >> > If so, shouldn't that be the focus of development rahter than >> > placing the burden on userspace apps to handle storage repair >> > situations? >> >> It really depends on the programming model. In the model Vishal is >> talking about, either the applications themselves or the libraries they >> link to are expected to implement the redundancies where necessary. > > IOWs, filesystems no longer have any control over data integrity. > Yet it's the filesystem developers who will still be responsible for > data integrity and when the filesystem has a data coruption event > we'll get blamed and the filesystem gets a bad name, even though > it's entirely the applications fault. We've seen this time and time > again - application developers cannot be trusted to guarantee data > integrity. yes, some apps will be fine, but do you really expect > application devs that refuse to use fsync because it's too slow are > going to have a different approach to integrity when it comes to > DAX? Yes, completely agree. The applications that will implement competent error recovery with these mechanisms will be vanishingly small, and there is definite room for a kernel data-redundancy solution that builds on these patches. > >> >> > There's an implicit assumption that applications will keep redundant >> >> > copies of their data at the /application layer/ and be able to >> >> > automatically repair it? >> >> That's one way to do things. It really depends on the application what >> it will do for recovery. >> >> >> > And then there's the implicit assumption that it will unlink and >> >> > free the entire file before writing a new copy >> >> I think Vishal was referring to restoring from backup. cp itself will >> truncate the file before overwriting, iirc. > > Which version of cp? what happens if they use --sparse and the error > is in a zeroed region? There's so many assumptions about undefined userspace > environment, application and user behaviour being made here, and > it's all being handwaved away. > > I'm asking for this to be defined, demonstrated and documented as a > working model that cannot be abused and doesn't have holes the size > of trucks in it, not handwaving... You lost me... how are these patches abusing the existing semantics of -EIO and write to clear? >> >> To summarize, the two cases we want to handle are: >> >> 1. Application has inbuilt recovery: >> >> - hits badblock >> >> - figures out it is able to recover the data >> >> - handles SIGBUS or EIO >> >> - does a (sector aligned) write() to restore the data >> > >> > The "figures out" step here is where >95% of the work we'd have to >> > do is. And that's in filesystem and block layer code, not >> > userspace, and userspace can't do that work in a signal handler. >> > And it can still fall down to the second case when the application >> > doesn't have another copy of the data somewhere. >> >> I read that "figures out" step as the application determining whether or >> not it had a redundant copy. > > Another undocumented assumption, that doesn't simplify what needs to > be done. Indeed, userspace can't do that until it is in SIGBUS > context, which tends to imply applications need to do a major amount > of work from within the signal handler.... > >> > FWIW, we don't have a DAX enabled filesystem that can do >> > reverse block mapping, so we're a year or two away from this being a >> > workable production solution from the filesystem perspective. And >> > AFAICT, it's not even on the roadmap for dm/md layers. >> >> Do we even need that? What if we added an FIEMAP flag for determining >> bad blocks. > > So you're assuming that the filesystem has been informed of the bad > blocks and has already marked the bad regions of the file in it's > extent list? > > How does that happen? What mechanism is used for the underlying > block device to inform the filesytem that theirs a bad LBA, and how > does the filesytem the map that to a path/file/offset with reverse > mapping? Or is there some other magic that hasn't been explained > happening here? In 4.5 we added this: commit 99e6608c9e7414ae4f2168df8bf8fae3eb49e41f Author: Vishal Verma <vishal.l.verma@intel.com> Date: Sat Jan 9 08:36:51 2016 -0800 block: Add badblock management for gendisks NVDIMM devices, which can behave more like DRAM rather than block devices, may develop bad cache lines, or 'poison'. A block device exposed by the pmem driver can then consume poison via a read (or write), and cause a machine check. On platforms without machine check recovery features, this would mean a crash. The block device maintaining a runtime list of all known sectors that have poison can directly avoid this, and also provide a path forward to enable proper handling/recovery for DAX faults on such a device. Use the new badblock management interfaces to add a badblocks list to gendisks. Signed-off-by: Vishal Verma <vishal.l.verma@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
On Mon, May 02, 2016 at 10:53:25AM -0700, Dan Williams wrote: > On Mon, May 2, 2016 at 8:18 AM, Jeff Moyer <jmoyer@redhat.com> wrote: > > Dave Chinner <david@fromorbit.com> writes: > [..] > >> We need some form of redundancy and correction in the PMEM stack to > >> prevent single sector errors from taking down services until an > >> administrator can correct the problem. I'm trying to understand > >> where this is supposed to fit into the picture - at this point I > >> really don't think userspace applications are going to be able to do > >> this reliably.... > > > > Not all storage is configured into a RAID volume, and in some instances, > > the application is better positioned to recover the data (gluster/ceph, > > for example). It really comes down to whether applications or libraries > > will want to implement redundancy themselves in order to get a bump in > > performance by not going through the kernel. And I think I know what > > your opinion is on that front. :-) > > > > Speaking of which, did you see the numbers Dan shared at LSF on how much > > overhead there is in calling into the kernel for syncing? Dan, can/did > > you publish that spreadsheet somewhere? > > Here it is: > > https://docs.google.com/spreadsheets/d/1pwr9psy6vtB9DOsc2bUdXevJRz5Guf6laZ4DaZlkhoo/edit?usp=sharing > > On the "Filtered" tab I have some of the comparisons where: Those numbers are really wacky - the inconsistent decimal place representation makes it really, really hard to read the differences in orders of magnitude, too. Let's take the first numbers - noop, 64 byte ops are: threads ops/s 1 90M 2 310M 4 65M 8 175M 16 426M Why aren't these linear? And if the test is not running in an environment where these are controlled and linear, how valid are the rest of the tests and hence the comparison. > noop => don't call msync and don't flush caches in userspace > > persist => cache flushing only in userspace and only on individual cache lines So these look a lot more linear than the no-op behaviour, so I'll just ignore the no-op results for now. > persist_4k => cache flushing only in userspace, but flushing is > performed in 4K aligned units Urg, your "vs persist" percentages are all wrong. You can't have a "-1000%" difference, you have "persist 4k" running at 10% of the speed of "persist". So, with that in mind, the "persist_4k" speed is: ops/s single thread Size vs "persist" 4k flush rate 64 10% 834k 128 13% 849k 256 15% 410k(one off variation?) 512 20% 860k 1024 25% 850k 2048 50% 840k 4096 none 836k 8192 none 410k What we see here is that the CPU(s) can flush the 4k pages at a rate of roughly 850,000 flushes/s, whilst the 64 byte flush rate is around 8.8M flushes/s. This is clearly demonstrated in the numbers - as the dirty object size approaches the cache flush granularity, the speed approaches single cacheline flush granularity speed. Comparing 4k vs 64b flushes, we have 63 clean cache line flushes taking roughly the same time as 9 dirty cache line flushes. Nice numbers - that means a clean cache line flush has ~14% of the overhead of dirty cache line flush. Seems rather high - it's tens of CPU cycles to determine that the flush is a no-op for that cacheline. Fixing this seems like a hardware optimisation issue to me, but I still have to question how many applications are going to have such fine-grained random synchronous memory writes that this actually matters in practice? If we are doing such small writes across multiple different 4k pages, then TLB overhead for all the page faults is going to be as much of an issue as 4k cache flushes... > msync => same granularity flushing as the 'persist' case, but the > kernel internally promotes this to a 4K sized / aligned flush So you're calling msync for every modification that is made? What application needs to do that? Anyway, page flush rates paint an interesting picture: single thread versus Size 4k flush rate persist_4k 64 655k 78% 128 655k 81% 256 670k 163% (* persist 4k number low) 512 681k 79% 1024 666k 78% 2048 650k 77% 4096 652k 78% 8192 390k 95% msync adds relatively little overhead (~20% extra overhead) compared to the performance loss from the 4k flush granularity change. And given this appears to be a worst case test scenario (and I'm sure msync could be improved), I don't think this demonstrates a problem with using msync. IMO, these numbers don't support the argument that the *msync model* for data integrity for DAX is flawed, unworkable, or too slow. What I see is a performance problem resulting from the overhead of flushing clean cachelines. i.e. there's data here that supports the argument for reducing the overhead of flushing clean cachelines in the hardware and/or better tracking of dirty cachelines within the kernel, but not data that says the msync() based data integrity model is the source of the problem. i.e. separate the programming model from the performance issue, and we can see that the performance problem is not caused by the programming model - it's caused by the kernel implementation of the model. > The takeaway is that msync() is 9-10x slower than userspace cache management. An alternative viewpoint: that flushing clean cachelines is extremely expensive on Intel CPUs. ;) i.e. Same numbers, different analysis from a different PoV, and that gives a *completely different conclusion*. Think about it for the moment. The hardware inefficiency being demonstrated could be fixed/optimised in the next hardware product cycle(s) and so will eventually go away. OTOH, we'll be stuck with whatever programming model we come up with for the next 30-40 years, and we'll never be able to fix flaws in it because applications will be depending on them. Do we really want to be stuck with a pmem model that is designed around the flaws and deficiencies of ~1st generation hardware? Cheers, Dave.
>> The takeaway is that msync() is 9-10x slower than userspace cache management. > >An alternative viewpoint: that flushing clean cachelines is >extremely expensive on Intel CPUs. ;) > >i.e. Same numbers, different analysis from a different PoV, and >that gives a *completely different conclusion*. > >Think about it for the moment. The hardware inefficiency being >demonstrated could be fixed/optimised in the next hardware product >cycle(s) and so will eventually go away. OTOH, we'll be stuck with >whatever programming model we come up with for the next 30-40 years, >and we'll never be able to fix flaws in it because applications will >be depending on them. Do we really want to be stuck with a pmem >model that is designed around the flaws and deficiencies of ~1st >generation hardware? Hi Dave, Not sure I agree with your completely different conclusion. (Not sure I completely disagree either, but please let me raise some practical points.) First of all, let's say you're completely right and flushing clean cache lines is extremely expensive. So your solution is to wait for the chip to be fixed? Remember the model we're putting forward (which we're working on documenting, because I fully agree with the lack of documentation point you keep raising) requires the application to ASK for the file system's permission before assuming flushing from user space to persistence is allowed. So that doesn't stick us with 30-40 years of a flawed model. I don't think the model is wrong, having spent lots of research time on it, but if I'm full of crap, all we have to do is stop telling the app that flushing from user space is allowed and it must go back to using msync(). This is my understanding of what Dan suggested at LSF and this is what I'm currently writing up. By the way, the NVM Libraries already contain the logic to ask if flushing from user space is allowed, falling back to msync() if not. Currently those libraries check for DAX mappings. But the points you raised about metadata changes happening during page faults made us realize we have to ask the file system to opt-in to allowing user space flushing, so that's what we're changing the library to do. See, we are listening :-) Anyway, I doubt that flushing a clean cache line is extremely expensive. Remember the code is building transactions to maintain a consistent in-memory data structure in the face of sudden failure like powerloss. So it is using the flushes to create store barriers, but not the block- based store barriers we're used to in the storage world, but cache-line- sized store barriers (usually multiples of cache lines, but most commonly smaller than 4k of them). So I think when you turn a cache line flush into an msync(), you're seeing some dirty stuff get flushed before it is time to flush it. I'm not sure though, but certainly we could spend more time testing & measuring. More importantly, I think it is interesting to decide what we want the pmem programming model to be long-term. I think we want applications to just map pmem, do normal stores to it, and assume they are persistent. This is quite different from the 30-year-old POSIX Model where msync() is required. But I think it is cleaner, easier to understand, and less error-prone. So why doesn't it work that way right now? Because we're finding it impractical. Using write-through caching for pmem simply doesn't perform well, and depending on the platform to flush the CPU caches on shutdown/powerfail is not practical yet. But I think the day will come when it is practical. So given that long-term target, the idea is for an application to ask if the msync() calls are required, or if just flushing the CPU caches is sufficient for persistence. Then, we're also adding an ACPI property that allows SW to discover if the caches are flushed automatically on shutdown/powerloss. Initially that will only be true for custom platforms, but hopefully it can be available more broadly in the future. The result will be that the programming model gets simpler as more and more hardware requires less explicit flushing. Now I'll go back to writing up the big picture for this programming model so I can ask you for comments on that as well... -andy
On Mon, May 02, 2016 at 04:25:51PM -0700, Dan Williams wrote: > On Mon, May 2, 2016 at 4:04 PM, Dave Chinner <david@fromorbit.com> wrote: > > On Mon, May 02, 2016 at 11:18:36AM -0400, Jeff Moyer wrote: > >> Dave Chinner <david@fromorbit.com> writes: > >> > >> > On Mon, Apr 25, 2016 at 11:53:13PM +0000, Verma, Vishal L wrote: > >> >> On Tue, 2016-04-26 at 09:25 +1000, Dave Chinner wrote: > >> > You're assuming that only the DAX aware application accesses it's > >> > files. users, backup programs, data replicators, fileystem > >> > re-organisers (e.g. defragmenters) etc all may access the files and > >> > they may throw errors. What then? > >> > >> I'm not sure how this is any different from regular storage. If an > >> application gets EIO, it's up to the app to decide what to do with that. > > > > Sure - they'll fail. But the question I'm asking is that if the > > application that owns the data is supposed to do error recovery, > > what happens when a 3rd party application hits an error? If that > > consumes the error, the the app that owns the data won't ever get a > > chance to correct the error. > > > > This is a minefield - a 3rd party app that swallows and clears DAX > > based IO errors is a data corruption vector. can yo imagine if > > *grep* did this? The model that is being promoted here effectively > > allows this sort of behaviour - I don't really think we > > should be architecting an error recovery strategy that has the > > capability to go this wrong.... > > Since when does grep write to a file on error? That's precisely my point - it doesn't right now because there is no onus on userspace applications to correct data errors when they are found. However, if the accepted model becomes "userspace needs to try to correct errors in data automatically", the above scenario is a distinct possiblity. I'm not saying grep will do this - I'm taking the logical argument being presented to the extreme - but I'm sure that there will be developers that have enough knowledge to know they are supposed to do something with errors on pmem devices, but not have enough knowledge to know the correct things to do. And then the app mishandles a EINVAL error (or something like that) and so we end up with buggy userspace apps trying to correct errors in good data and causing data loss that way. Do we really want to introduce a data integrity and error recovery model where this sort of "bug" is a distinct possibly? > >> >> > There's an implicit assumption that applications will keep redundant > >> >> > copies of their data at the /application layer/ and be able to > >> >> > automatically repair it? > >> > >> That's one way to do things. It really depends on the application what > >> it will do for recovery. > >> > >> >> > And then there's the implicit assumption that it will unlink and > >> >> > free the entire file before writing a new copy > >> > >> I think Vishal was referring to restoring from backup. cp itself will > >> truncate the file before overwriting, iirc. > > > > Which version of cp? what happens if they use --sparse and the error > > is in a zeroed region? There's so many assumptions about undefined userspace > > environment, application and user behaviour being made here, and > > it's all being handwaved away. > > > > I'm asking for this to be defined, demonstrated and documented as a > > working model that cannot be abused and doesn't have holes the size > > of trucks in it, not handwaving... > > You lost me... how are these patches abusing the existing semantics > of -EIO and write to clear? I haven't said that. I said there are assumptions about how userspace will handle the error, but they aren't documented anywhere. "copy a file using cp" is not a robust recovery solution - it provides no guarantees about how the bad file and regions will be recycled and the errors cleared. This effectively of puts it all on the filesystems to deal with, even though you're trying to design an error handling model that bypasses the filesystems and goes straight to userspace. If I can't understand how this is all supposed to work because none of it is documented, then we have no chance that the average admin is going to be able to understand it. > > >> >> To summarize, the two cases we want to handle are: > >> >> 1. Application has inbuilt recovery: > >> >> - hits badblock > >> >> - figures out it is able to recover the data > >> >> - handles SIGBUS or EIO > >> >> - does a (sector aligned) write() to restore the data > >> > > >> > The "figures out" step here is where >95% of the work we'd have to > >> > do is. And that's in filesystem and block layer code, not > >> > userspace, and userspace can't do that work in a signal handler. > >> > And it can still fall down to the second case when the application > >> > doesn't have another copy of the data somewhere. > >> > >> I read that "figures out" step as the application determining whether or > >> not it had a redundant copy. > > > > Another undocumented assumption, that doesn't simplify what needs to > > be done. Indeed, userspace can't do that until it is in SIGBUS > > context, which tends to imply applications need to do a major amount > > of work from within the signal handler.... > > > >> > FWIW, we don't have a DAX enabled filesystem that can do > >> > reverse block mapping, so we're a year or two away from this being a > >> > workable production solution from the filesystem perspective. And > >> > AFAICT, it's not even on the roadmap for dm/md layers. > >> > >> Do we even need that? What if we added an FIEMAP flag for determining > >> bad blocks. > > > > So you're assuming that the filesystem has been informed of the bad > > blocks and has already marked the bad regions of the file in it's > > extent list? > > > > How does that happen? What mechanism is used for the underlying > > block device to inform the filesytem that theirs a bad LBA, and how > > does the filesytem the map that to a path/file/offset with reverse > > mapping? Or is there some other magic that hasn't been explained > > happening here? > > In 4.5 we added this: > > commit 99e6608c9e7414ae4f2168df8bf8fae3eb49e41f > Author: Vishal Verma <vishal.l.verma@intel.com> > Date: Sat Jan 9 08:36:51 2016 -0800 > > block: Add badblock management for gendisks Yes, I know, and it doesn't answer any of the questions I just asked. What you just told me is that there is something that is kept three levels of abstraction away from a filesystem. So: - What mechanism is to be used for the underlying block device to inform the filesytem that a new bad block was added to this list? What context comes along with that notification? - how does the filesystem query the bad block list without adding layering violations? - when does the filesystem need to query the bad block list? - how will the bad block list propagate through DM/MD layers? - how does the filesytem the map the bad block to a path/file/offset without reverse mapping - does this error handling interface really imply the filesystem needs to implement brute force scans at notification time? - Is the filesystem expectd to find the active application or address_space access that triggered the bad block notification to handle them correctly? (e.g. prevent a page fault from failing because we can recover from the error immediately) - what exactly is the filesystem supposed to do with the bad block? e.g: - is the block persistently bad until the filesystem rewrites it? Over power cycles? Will we get multiple notifications (e.g. once per boot)? - Is the filesystem supposed to intercept reads/writes to bad blocks once it knows about them? - how is the filesystem supposed to communicate that there is a bad block in a file back to userspace? Or is userspace supposed to infer that there's a bad block from EIO and so has to run FIEMAP to determine if the error really was due to a bad block? - what happens if there is no running application that we can report the error to or will handle the error (e.g. found error by a media scrub or during boot)? - if the bad block is in filesystem free space, what should the filesystem do with it? What I'm failing to communicate is that having and maintaining things like bad block lists in a block device is the easy part of the problem. Similarly reporting a bad block flag in FIEMAP is only a few lines of code to implement, but that assumes the filesystem has already propagated the bad block information into it's internal extents lists. That's the hard part of all this: connecting the two pieces together in a sane, reliable, consistent and useful manner. This will form the user API, so we need to sort it out before applications start to use it. However, if I'm struggling to understand how I'm supposed to connecct up the parts inside a filesytem, then expecting application developers to be able to connect the dots in a sane manner is bordering on fantasy.... Cheers, Dave.
On Tue, May 03, 2016 at 01:26:46AM +0000, Rudoff, Andy wrote: > > >> The takeaway is that msync() is 9-10x slower than userspace cache management. > > > >An alternative viewpoint: that flushing clean cachelines is > >extremely expensive on Intel CPUs. ;) > > > >i.e. Same numbers, different analysis from a different PoV, and > >that gives a *completely different conclusion*. > > > >Think about it for the moment. The hardware inefficiency being > >demonstrated could be fixed/optimised in the next hardware product > >cycle(s) and so will eventually go away. OTOH, we'll be stuck with > >whatever programming model we come up with for the next 30-40 years, > >and we'll never be able to fix flaws in it because applications will > >be depending on them. Do we really want to be stuck with a pmem > >model that is designed around the flaws and deficiencies of ~1st > >generation hardware? > > Hi Dave, > > Not sure I agree with your completely different conclusion. (Not sure > I completely disagree either, but please let me raise some practical > points.) > > First of all, let's say you're completely right and flushing clean > cache lines is extremely expensive. So your solution is to wait for > the chip to be fixed? No, I'm not saying that's the solution - I'm pointing out that if clean cache line flushing overhead is less of a problem in future, the optimisations made now will not be necessary. However, we'll be still stuck with the API/model that has encoded those optimisations as a necessary thing for applications to know about and do the correct thing with. I.e. we end up with a library of applications that are optimised for a problem that no longer exists... > Remember the model we're putting forward (which > we're working on documenting, because I fully agree with the lack of > documentation point you keep raising) requires the application to ASK > for the file system's permission before assuming flushing from user space > to persistence is allowed. And when the filesystem says no because the fs devs don't want to have to deal with broken apps because app devs learn that "this is a go fast knob" and data integrity be damned? It's "fsync is slow so I won't use it" all over again... > Anyway, I doubt that flushing a clean cache line is extremely expensive. > Remember the code is building transactions to maintain a consistent > in-memory data structure in the face of sudden failure like powerloss. > So it is using the flushes to create store barriers, but not the block- > based store barriers we're used to in the storage world, but cache-line- > sized store barriers (usually multiples of cache lines, but most commonly > smaller than 4k of them). So I think when you turn a cache line flush > into an msync(), you're seeing some dirty stuff get flushed before it > is time to flush it. I'm not sure though, but certainly we could spend > more time testing & measuring. Sure, but is that what Dan was testing? I don't know - he just presented a bunch of numbers without a description of the workload, posting the benchmark code, etc. hence I can only *make assumptions* about what the numbers mean. I'm somewhat tired of having to make assumptions because nobody is describing what they are doing sufficiently and then getting called out for it, or having to ask lots of questions because other people have made assumptions about how they think something is going to work without explaining how the dots connect together. It's a waste of everyone's time to be playing this ass-u-me game... The fact that nobody has been able to explain the how the overall model is supposed to work from physical error all the way out to userspace makes me think that this is all being made up on the spot. There are big pieces of the picture missing, and nobody seems to be able to communicate a clear vision of the architecture we are supposed to be discussing, let alone implementing... > More importantly, I think it is interesting to decide what we want the > pmem programming model to be long-term. I think we want applications to > just map pmem, do normal stores to it, and assume they are persistent. > This is quite different from the 30-year-old POSIX Model where msync() > is required. Yes, it's different, but we still have to co-ordinate multiple layers of persistence (i.e. metadata that references the data). > But I think it is cleaner, easier to understand, and less > error-prone. So why doesn't it work that way right now? Because we're > finding it impractical. Using write-through caching for pmem simply > doesn't perform well, and depending on the platform to flush the CPU > caches on shutdown/powerfail is not practical yet. But I think the day > will come when it is practical. Right - it's also simply not practical to intercept every userspace store to ensure the referencing metadata is also persistent, so we still need synchronisation mechanisms to ensure that such state is acheived. Either that, or the entire dynamic filesystem state needs to be stored in write-through persistent memory as well. We're a long, long way from that. And, please keep in mind: many application developers will not design for pmem because they also have to support traditional storage backed by page cache. If they use msync(), the app will work on any storage stack, but just be much, much faster on pmem+DAX. So, really, we have to make the msync()-only model work efficiently, so we may as well design for that in the first place.... > So given that long-term target, the idea is for an application to ask if > the msync() calls are required, or if just flushing the CPU caches is > sufficient for persistence. Then, we're also adding an ACPI property > that allows SW to discover if the caches are flushed automatically > on shutdown/powerloss. Initially that will only be true for custom > platforms, but hopefully it can be available more broadly in the future. > The result will be that the programming model gets simpler as more and > more hardware requires less explicit flushing. That's a different problem, and one that requires a filesystem to also store all it's dynamic information in pmem. i.e. there's not point flushing pmem caches if the powerloss loses dirty metadata that is held in system RAM. We really need completely new pmem-native filesystems to make this work - it's a completely separate problem to whether msync() should be the API that provided fundamental data integrity guarantees or not. Which brings up another point: advanced new functionality is going to require native pmem filesystems. These are unlikely to be block device based, and instead will directly interface with the low level CPU and pmem APIs. I don't expect these to use the DAX infrastructure, either, because that assumes block device based operations. The will, however, still have to have POSIX compatible behaviour, and so we go full circle in expecting that an app written for mmap+DAX on an existing block based filesystem will work identically on funky new byte-addressable native pmem filesytems. Encoding cache flushing for data integrity into the userspace applications assumes that such future pmem-based storage will have identical persistence requirements to the existing hardware. This, to me, seems very unlikely to be the case (especially when considering different platforms (e.g. power, ARM)) and so, again, application developers are likely to have to fall back to using a kernel provided data integrity primitive they know they can rely on (i.e. msync())..... Cheers, Dave.
On Mon, May 2, 2016 at 6:51 PM, Dave Chinner <david@fromorbit.com> wrote: > On Mon, May 02, 2016 at 04:25:51PM -0700, Dan Williams wrote: [..] > Yes, I know, and it doesn't answer any of the questions I just > asked. What you just told me is that there is something that is kept > three levels of abstraction away from a filesystem. So: Ok, let's answer them. A lot of your questions seem to assume the filesystem has a leading role to play with error recovery, that isn't the case with traditional disk errors and we're not looking to change that situation. The filesystem can help with forensics after an error escapes the kernel and is communicated to userspace, but the ability to reverse map a sector to a file is just a convenience to identify potential data loss. For redundancy in the DAX case I can envision DAX-aware RAID that makes the potential exposure to bad blocks smaller, but it will always be the case that the array can be out-of-sync / degraded and has no choice but to communicate the error to userspace. So, the answers below address what we do when we are in that state, and include some thoughts about follow-on enabling we can do at the DM/MD layer. > - What mechanism is to be used for the underlying block > device to inform the filesytem that a new bad block was > added to this list? The filesystem doesn't need this notification and doesn't get it today from RAID. It's handy for the bad block list to be available to fs/dax.c and the block layer, but I don't see ext4/xfs having a role to play with the list and certainly not care about "new error detected events". For a DM/MD driver it also does not need to know about new errors because it will follow the traditional disk model where errors are handled on access, or discovered and scrubbed during a periodic array scan. That said, new errors may get added to the list by having the pmem driver trigger a rescan of the device whenever a latent error is discovered (i.e. memcpy_from_pmem() returns -EIO). The update of the bad block list is asynchronous. We also have a task on the todo list to allow the pmem rescan action to be triggered via sysfs. > What context comes along with that > notification? The only notification the file system gets is -EIO on access. However, assuming we had a DAX-aware RAID driver what infrastructure would we need to prevent SIGBUS from reaching the application if we happened to have a redundant copy of the data? One feature we've talked about for years at LSF/MM but never made any progress on is a way for a file system to discover and query if the storage layer can reconstruct data from redundant information. Assuming we had such an interface there's still the matter of plumbing a machine check fault through a physical-address-to-sector conversion and request the block device driver to attempt to provide a redundant copy. The in-kernel recovery path, assuming RAID is present, needs more thought especially considering the limited NMI context of a machine check notification and the need to trap back into driver code. I see the code in fs/dax.c getting involved to translate a process-physical-address back to a sector, but otherwise the rest of the filesystem need not be involved. > - how does the filesystem query the bad block list without > adding layering violations? Why does the file system need to read the list? Apologies for answering this question with a question, but these patches don't assume the filesystem will do anything with a bad block list. > - when does the filesystem need to query the bad block list? > - how will the bad block list propagate through DM/MD > layers? > - how does the filesytem the map the bad block to a > path/file/offset without reverse mapping - does this error > handling interface really imply the filesystem needs to > implement brute force scans at notification time? No, there is no implication that reverse mapping is a requirement. > - Is the filesystem expectd to find the active application or > address_space access that triggered the bad block > notification to handle them correctly? (e.g. prevent a > page fault from failing because we can recover from the > error immediately) With these patches no, but it would be nice to incrementally add that ability. I.e. trap machine check faults on non-anonymous memory and send a request down the stack to recover the sector if the storage layer has a redundant copy. Again, fs/dax.c would need extensions to do this coordination, but I don't foresee the filesystem getting involved beyond that point. > - what exactly is the filesystem supposed to do with the bad > block? e.g: > - is the block persistently bad until the filesystem > rewrites it? Over power cycles? Will we get > multiple notifications (e.g. once per boot)? Bad blocks on persistent memory media remain bad after a reboot. Per the ACPI spec the DIMM device tracks the errors and reports them in response to an "address range scrub" command. Each boot the libnvdimm sub-system kicks off a scrub and populates the bad block list per pmem namespace. As mentioned above, we want to add the ability to re-trigger this scrub on-demand, in response to a memcpy_from_pmem() discovering an error, or after a SIGBUS is communicated to userspace. > - Is the filesystem supposed to intercept > reads/writes to bad blocks once it knows about > them? No, the driver handles that. > - how is the filesystem supposed to communicate that > there is a bad block in a file back to userspace? -EIO on access. > Or is userspace supposed to infer that there's a > bad block from EIO and so has to run FIEMAP to > determine if the error really was due to a bad > block? The information is there to potentially do forensics on why an I/O encountered an error, but there is no expectation that userspace follow up on each -EIO with a FIEMAP. > - what happens if there is no running application > that we can report the error to or will handle the > error (e.g. found error by a media scrub or during > boot)? Same as RAID today, if the array is in sync the bad block will get re-written during the scrub hopefully in advance of when an application might discover it. If no RAID is present then the only notification is an on-access error. > - if the bad block is in filesystem free space, what should > the filesystem do with it? Nothing. When the free space becomes allocated we rely on the fact that the filesystem will first zero the blocks. That zeroing process will clear the media error. Now, in rare cases clearing the error might itself fail, but in that case we just degenerate to the latent error discovery case. > What I'm failing to communicate is that having and maintaining > things like bad block lists in a block device is the easy part of > the problem. > > Similarly reporting a bad block flag in FIEMAP is only a few > lines of code to implement, but that assumes the filesystem has > already propagated the bad block information into it's internal > extents lists. > > That's the hard part of all this: connecting the two pieces together > in a sane, reliable, consistent and useful manner. This will form > the user API, so we need to sort it out before applications start to > use it. Applications are already using most of this model today. The new enabling we should consider is a way to take advantage of redundancy at the storage driver layer to prevent errors from being reported to userspace when redundant data is available. Preventing machine-check-SIGBUS signals from reaching applications is a new general purpose error handling mechanism that might also be useful for DRAM errors outside of pmem+DAX. > However, if I'm struggling to understand how I'm supposed to > connecct up the parts inside a filesytem, then expecting application > developers to be able to connect the dots in a sane manner is > bordering on fantasy.... Hopefully it is becoming clearer that we are not proposing anything radically different than what is present for error recovery today modulo thinking about the mechanisms to trap and recover a DAX read of a bad media area via a DM/MD implementation. DAX writes on the other hand are more challenging in that we'd likely want to stage them and wait to commit them until an explicit sync point. However, this is still consistent with the DAX programming model. An application is free to trade off raw access to the media for higher-order filesystem and storage layer features provided by the kernel.
> >And when the filesystem says no because the fs devs don't want to >have to deal with broken apps because app devs learn that "this is a >go fast knob" and data integrity be damned? It's "fsync is slow so I >won't use it" all over again... ... > >And, please keep in mind: many application developers will not >design for pmem because they also have to support traditional >storage backed by page cache. If they use msync(), the app will work >on any storage stack, but just be much, much faster on pmem+DAX. So, >really, we have to make the msync()-only model work efficiently, so >we may as well design for that in the first place.... Both of these snippets seem to be arguing that we should make msync/fsync more efficient. But I don't think anyone is arguing the opposite. Is someone saying we shouldn't make the msync()-only model work efficiently? Said another way: the common case for DAX will be applications simply following the POSIX model. open, mmap, msync... That will work fine and of course we should optimize that path as much as possible. Less common are latency-sensitive applications built to leverage to byte- addressable nature of pmem. File systems supporting this model will indicate it using a new ioctl that says doing CPU cache flushes is sufficient to flush stores to persistence. But I don't see how that direction is getting turned into an argument against msync() efficiency. >Which brings up another point: advanced new functionality >is going to require native pmem filesystems. I agree there's opportunity for new filesystems (and old) to leverage what pmem provides. But the word "require" implies that's the only way to go and we know that's not the case. Using ext4+dax to map pmem into an application allows that application to use the pmem directly and a good number of software projects are doing exactly that. -andy
On Tue, May 03, 2016 at 06:30:04PM +0000, Rudoff, Andy wrote: > > > >And when the filesystem says no because the fs devs don't want to > >have to deal with broken apps because app devs learn that "this is a > >go fast knob" and data integrity be damned? It's "fsync is slow so I > >won't use it" all over again... > ... > > > >And, please keep in mind: many application developers will not > >design for pmem because they also have to support traditional > >storage backed by page cache. If they use msync(), the app will work > >on any storage stack, but just be much, much faster on pmem+DAX. So, > >really, we have to make the msync()-only model work efficiently, so > >we may as well design for that in the first place.... > > Both of these snippets seem to be arguing that we should make msync/fsync > more efficient. But I don't think anyone is arguing the opposite. Is > someone saying we shouldn't make the msync()-only model work efficiently? Not directly. The argument presented is "we need a flag to avoid msync, because msync is inefficient", which is followed by "look, here's numbers that show msync() being slow, so just give us the flag already". Experience tells me that the moment a workaround is in place, nobody will go back and try to fix the problem that the workaround is mitigating. Now we know that it's the page granularity cache flushing overhead that causes the performance differential rather than it being caused by using msync(), we should be looking at ways to reduce the cache flushing overhead, not completely bypassing it. > Said another way: the common case for DAX will be applications simply > following the POSIX model. open, mmap, msync... That will work fine > and of course we should optimize that path as much as possible. Less > common are latency-sensitive applications built to leverage to byte- > addressable nature of pmem. File systems supporting this model will > indicate it using a new ioctl that says doing CPU cache flushes is > sufficient to flush stores to persistence. You keep saying this whilst ignoring the repeated comments about how this can not be guaranteed by all filesystems, and hence apps will not be able to depend on having such behaviour present. The only guarantee for persistence that an app will be able to rely on is msync(). > But I don't see how that > direction is getting turned into an argument against msync() efficiency. Promoting a model that works around inefficiency rather than solving it is no different to saying you don't care about fixing the inefficiency.... I've said my piece, I'm not going to waste any more time going around this circle again. Cheers, Dave.
On Tue, May 03, 2016 at 10:28:15AM -0700, Dan Williams wrote: > On Mon, May 2, 2016 at 6:51 PM, Dave Chinner <david@fromorbit.com> wrote: > > On Mon, May 02, 2016 at 04:25:51PM -0700, Dan Williams wrote: > [..] > > Yes, I know, and it doesn't answer any of the questions I just > > asked. What you just told me is that there is something that is kept > > three levels of abstraction away from a filesystem. So: > > Ok, let's answer them. > > A lot of your questions seem to assume the filesystem has a leading > role to play with error recovery, that isn't the case with traditional > disk errors and we're not looking to change that situation. *cough* BTRFS New filesystems are mostly being designed with redundancy and recovery mechanisms built into them. Hence the high level /assumption/ that filesystems aren't going to play a significant role in error recovery for pmem storage is, well, somewhat disturbing.... > The > filesystem can help with forensics after an error escapes the kernel > and is communicated to userspace, but the ability to reverse map a > sector to a file is just a convenience to identify potential data > loss. So working out what file got corrupted in your terabytes of pmem storage is "just a convenience"? I suspect that a rather large percentage of admins will disagree with you on this. > For redundancy in the DAX case I can envision DAX-aware RAID that > makes the potential exposure to bad blocks smaller, but it will always > be the case that the array can be out-of-sync / degraded and has no > choice but to communicate the error to userspace. So, the answers > below address what we do when we are in that state, and include some > thoughts about follow-on enabling we can do at the DM/MD layer. > > > - What mechanism is to be used for the underlying block > > device to inform the filesytem that a new bad block was > > added to this list? > > The filesystem doesn't need this notification and doesn't get it today > from RAID. Why doesn't the filesystem need this notification? Just because we don't get it today from a RAID device does not mean we can't use it. Indeed, think about the btrfs scrub operation - it validates everything on it's individual block devices, and when it finds a problem (e.g. a data CRC error) it notifies a different layer in the btrfs code that goes and works out if it can repair the problem from redundant copies/parity/mirrors/etc. > It's handy for the bad block list to be available to > fs/dax.c and the block layer, but I don't see ext4/xfs having a role > to play with the list and certainly not care about "new error detected > events". That's very short-sighted. Just because ext4/xfs don't *currently* do this, it doesn't mean other filesystems (existing or new) can't make use of notifications, nor that ext4/XFS can't ever make use of it, either. > For a DM/MD driver it also does not need to know about new > errors because it will follow the traditional disk model where errors > are handled on access, or discovered and scrubbed during a periodic > array scan. > > That said, new errors may get added to the list by having the pmem > driver trigger a rescan of the device whenever a latent error is > discovered (i.e. memcpy_from_pmem() returns -EIO). The update of the > bad block list is asynchronous. We also have a task on the todo list > to allow the pmem rescan action to be triggered via sysfs. IOWs, the pmem driver won't report errors to anyone who can correct them until an access to that bad block is made? Even if it means the error might go unreported and hence uncorrected for weeks or months because no access is made to that bad data? > > What context comes along with that > > notification? > > The only notification the file system gets is -EIO on access. > However, assuming we had a DAX-aware RAID driver what infrastructure > would we need to prevent SIGBUS from reaching the application if we > happened to have a redundant copy of the data? We'd need the same infrastructure at the filesystem layer would require if it has a redundant copy of the data. I don't know what that is, however, because I know very little about about MCEs and signal delivery (which is why I asked this question). [....] > The in-kernel recovery path, assuming RAID is present, needs more > thought especially considering the limited NMI context of a machine > check notification and the need to trap back into driver code. This is precisely the problem I am asking about - I know there is a limited context, but how exactly is it limited and what can we actually do from this context? e.g. Can we schedule recovery work on other CPU cores and wait for it to complete in a MCE notification handler? > I see > the code in fs/dax.c getting involved to translate a > process-physical-address back to a sector, but otherwise the rest of > the filesystem need not be involved. More /assumptions/ about filesystems not containing or being able to recover from redudant copies of data. > > > - how does the filesystem query the bad block list without > > adding layering violations? > > Why does the file system need to read the list? > Apologies for answering this question with a question, but these > patches don't assume the filesystem will do anything with a bad block > list. People keep talking about FIEMAP reporting bad blocks in files! How the fuck are we supposed to report bad blocks in a file via FIEMAP if the filesystem can't access the bad block list? > > - Is the filesystem expectd to find the active application or > > address_space access that triggered the bad block > > notification to handle them correctly? (e.g. prevent a > > page fault from failing because we can recover from the > > error immediately) > > With these patches no, but it would be nice to incrementally add that > ability. I.e. trap machine check faults on non-anonymous memory and > send a request down the stack to recover the sector if the storage > layer has a redundant copy. Again, fs/dax.c would need extensions to > do this coordination, but I don't foresee the filesystem getting > involved beyond that point. Again, the /assumption/ here is that only the block layer has the ability to recover, and only sector mapping is required from the fs. > > - what exactly is the filesystem supposed to do with the bad > > block? e.g: > > - is the block persistently bad until the filesystem > > rewrites it? Over power cycles? Will we get > > multiple notifications (e.g. once per boot)? > > Bad blocks on persistent memory media remain bad after a reboot. Per > the ACPI spec the DIMM device tracks the errors and reports them in > response to an "address range scrub" command. Each boot the libnvdimm > sub-system kicks off a scrub and populates the bad block list per pmem > namespace. As mentioned above, we want to add the ability to > re-trigger this scrub on-demand, in response to a memcpy_from_pmem() > discovering an error, or after a SIGBUS is communicated to userspace. > > > - Is the filesystem supposed to intercept > > reads/writes to bad blocks once it knows about > > them? > > No, the driver handles that. So, -EIO will be returned to the filesystem on access? If -EIO, then we'll have to check over the bad block list to determine if data recovery operations are required, right? Perhaps we need a different error here to tell the higher layers it's a specific type of error (e.g. -EBADBLOCK)? > > - how is the filesystem supposed to communicate that > > there is a bad block in a file back to userspace? > > -EIO on access. So no consideration for proactive "data loss has occurred at offset X in file /mnt/path/to/file, attempting recovery" messages when the error is first detected by the lowest layers? > > Or is userspace supposed to infer that there's a > > bad block from EIO and so has to run FIEMAP to > > determine if the error really was due to a bad > > block? > > The information is there to potentially do forensics on why an I/O > encountered an error, but there is no expectation that userspace > follow up on each -EIO with a FIEMAP. Ok, so how is userspace driven error recovery supposed to work if it can't differentiate the cause of an EIO error? If there's no requirement for FIEMAP to report the bad blocks in a file that needs recovery, then what is the app supposed to do with the EIO? Indeed, what consideration has been given to ensuring the app knows aheadi of time that the filesystem FIEMAP implementation will report bad blocks if they exist? Of course, the filesystem has to know about the bad blocks to be able to do any of this with FIEMAP.... > > - what happens if there is no running application > > that we can report the error to or will handle the > > error (e.g. found error by a media scrub or during > > boot)? > > Same as RAID today, if the array is in sync the bad block will get > re-written during the scrub hopefully in advance of when an > application might discover it. If no RAID is present then the only > notification is an on-access error. More /assumptions/ that only device level RAID will be able to recover.... > > - if the bad block is in filesystem free space, what should > > the filesystem do with it? > > Nothing. When the free space becomes allocated we rely on the fact > that the filesystem will first zero the blocks. That zeroing process > will clear the media error. The incorrect /assumption/ here is that all allocations will do block zeroing first. That's simply wrong. We do that for *user data* in XFS and ext4, but we do not do it for metadata as they are not accessed by DAX and, being transactionally protected, don't need zeroing to prevent stale data exposure. Hence we have a problem here - the first write to such blocks may be metadata writeback of some type and so the filesystem will see EIO errors in metadata writes and they'll freak out. What now - does this really mean that we'll have to add special IO falback code for all internal IO paths to be able to clear pmem bad block errors? Oh, and just a thought: lots of people are pushing for selectable FALLOC_FL_NO_HIDE_STALE behaviour which will skip zeroing of data blocks on allocation. If this happens, it we also skip the zeroing on allocation, so again there is no mechanism to clear bad block status in this case..... [...] > > However, if I'm struggling to understand how I'm supposed to > > connecct up the parts inside a filesytem, then expecting application > > developers to be able to connect the dots in a sane manner is > > bordering on fantasy.... > > Hopefully it is becoming clearer that we are not proposing anything > radically different than what is present for error recovery today > modulo thinking about the mechanisms to trap and recover a DAX read of > a bad media area via a DM/MD implementation. There is a radical difference - there is a pervasive /assumption/ in what is being proposed that filesystems are incapable of storing redundant information that can be used for error recovery. The current IO stack makes no such assumptions, even if it doesn't provide any infrastructure for such functionality. Cheers, Dave.
On Tue, May 3, 2016 at 8:18 PM, Dave Chinner <david@fromorbit.com> wrote: > On Tue, May 03, 2016 at 10:28:15AM -0700, Dan Williams wrote: >> On Mon, May 2, 2016 at 6:51 PM, Dave Chinner <david@fromorbit.com> wrote: >> > On Mon, May 02, 2016 at 04:25:51PM -0700, Dan Williams wrote: >> [..] >> > Yes, I know, and it doesn't answer any of the questions I just >> > asked. What you just told me is that there is something that is kept >> > three levels of abstraction away from a filesystem. So: >> >> Ok, let's answer them. >> >> A lot of your questions seem to assume the filesystem has a leading >> role to play with error recovery, that isn't the case with traditional >> disk errors and we're not looking to change that situation. > > *cough* BTRFS > > New filesystems are mostly being designed with redundancy and > recovery mechanisms built into them. Hence the high level > /assumption/ that filesystems aren't going to play a significant > role in error recovery for pmem storage is, well, somewhat > disturbing.... It is unfortunate is that you cite the lack of pmem enabling in btrfs as a reason to block patches that hookup the kernel's existing error mechanisms for the DAX case. I expect btrfs multi-device redundancy-management for pmem to be a more a coherent solution than what we can achieve with single-device-filesystems + RAID. I'm trying not to boil the ocean in this discussion, but Iet's go ahead and rope in btrfs-devel into this thread so we can make progress on hooking up SIGBUS notifications for DAX errors and bypassing dax_do_io() to clear errors. >> The >> filesystem can help with forensics after an error escapes the kernel >> and is communicated to userspace, but the ability to reverse map a >> sector to a file is just a convenience to identify potential data >> loss. > > So working out what file got corrupted in your terabytes of pmem > storage is "just a convenience"? I suspect that a rather large > percentage of admins will disagree with you on this. Yes, I will point them to their file system maintainer to ask about reverse mapping support. >> For redundancy in the DAX case I can envision DAX-aware RAID that >> makes the potential exposure to bad blocks smaller, but it will always >> be the case that the array can be out-of-sync / degraded and has no >> choice but to communicate the error to userspace. So, the answers >> below address what we do when we are in that state, and include some >> thoughts about follow-on enabling we can do at the DM/MD layer. >> >> > - What mechanism is to be used for the underlying block >> > device to inform the filesytem that a new bad block was >> > added to this list? >> >> The filesystem doesn't need this notification and doesn't get it today >> from RAID. > > Why doesn't the filesystem need this notification? Just because we > don't get it today from a RAID device does not mean we can't use it. If xfs and ext4 had a use for error notification today we would hook into it. > Indeed, think about the btrfs scrub operation - it validates > everything on it's individual block devices, and when it finds a > problem (e.g. a data CRC error) it notifies a different layer in the > btrfs code that goes and works out if it can repair the problem from > redundant copies/parity/mirrors/etc. Yes, just like RAID, sounds like we should definitely keep that in mind for the patch set that adds pmem support to btrfs, this isn't that patch set. >> It's handy for the bad block list to be available to >> fs/dax.c and the block layer, but I don't see ext4/xfs having a role >> to play with the list and certainly not care about "new error detected >> events". > > That's very short-sighted. Just because ext4/xfs don't *currently* > do this, it doesn't mean other filesystems (existing or new) can't > make use of notifications, nor that ext4/XFS can't ever make use of > it, either. Did I say "can't ever make use of it", no, if you have a need for a notification for xfs let's work on a notification mechanism. > >> For a DM/MD driver it also does not need to know about new >> errors because it will follow the traditional disk model where errors >> are handled on access, or discovered and scrubbed during a periodic >> array scan. >> >> That said, new errors may get added to the list by having the pmem >> driver trigger a rescan of the device whenever a latent error is >> discovered (i.e. memcpy_from_pmem() returns -EIO). The update of the >> bad block list is asynchronous. We also have a task on the todo list >> to allow the pmem rescan action to be triggered via sysfs. > > IOWs, the pmem driver won't report errors to anyone who can correct > them until an access to that bad block is made? Even if it means the > error might go unreported and hence uncorrected for weeks or months > because no access is made to that bad data? RAID periodically polls for and fixes bad blocks. The currently implementation only polls for errors at driver load. When we implement userspace triggered bad blocks scans we could also have a cron job to periodically kick off a scan, which follows the status quo for RAID error scanning. > >> > What context comes along with that >> > notification? >> >> The only notification the file system gets is -EIO on access. >> However, assuming we had a DAX-aware RAID driver what infrastructure >> would we need to prevent SIGBUS from reaching the application if we >> happened to have a redundant copy of the data? > > We'd need the same infrastructure at the filesystem layer would > require if it has a redundant copy of the data. I don't know what > that is, however, because I know very little about about MCEs and > signal delivery (which is why I asked this question). Fair enough. > [....] > >> The in-kernel recovery path, assuming RAID is present, needs more >> thought especially considering the limited NMI context of a machine >> check notification and the need to trap back into driver code. > > This is precisely the problem I am asking about - I know there is a > limited context, but how exactly is it limited and what can we > actually do from this context? e.g. Can we schedule recovery work on > other CPU cores and wait for it to complete in a MCE notification > handler? > >> I see >> the code in fs/dax.c getting involved to translate a >> process-physical-address back to a sector, but otherwise the rest of >> the filesystem need not be involved. > > More /assumptions/ about filesystems not containing or being able to > recover from redudant copies of data. > >> >> > - how does the filesystem query the bad block list without >> > adding layering violations? >> >> Why does the file system need to read the list? >> Apologies for answering this question with a question, but these >> patches don't assume the filesystem will do anything with a bad block >> list. > > People keep talking about FIEMAP reporting bad blocks in files! How > the fuck are we supposed to report bad blocks in a file via FIEMAP > if the filesystem can't access the bad block list? Compare FIEMAP results against the badblocks list, or arrange for FIEMAP to return the errors in file list by parsing the list against the badblocks list available in the gendisk. Shall I send a patch? It does assume we can do a inode to bdev lookup. >> > - Is the filesystem expectd to find the active application or >> > address_space access that triggered the bad block >> > notification to handle them correctly? (e.g. prevent a >> > page fault from failing because we can recover from the >> > error immediately) >> >> With these patches no, but it would be nice to incrementally add that >> ability. I.e. trap machine check faults on non-anonymous memory and >> send a request down the stack to recover the sector if the storage >> layer has a redundant copy. Again, fs/dax.c would need extensions to >> do this coordination, but I don't foresee the filesystem getting >> involved beyond that point. > > Again, the /assumption/ here is that only the block layer has the > ability to recover, and only sector mapping is required from the fs. Presently yes, in the future, no. > >> > - what exactly is the filesystem supposed to do with the bad >> > block? e.g: >> > - is the block persistently bad until the filesystem >> > rewrites it? Over power cycles? Will we get >> > multiple notifications (e.g. once per boot)? >> >> Bad blocks on persistent memory media remain bad after a reboot. Per >> the ACPI spec the DIMM device tracks the errors and reports them in >> response to an "address range scrub" command. Each boot the libnvdimm >> sub-system kicks off a scrub and populates the bad block list per pmem >> namespace. As mentioned above, we want to add the ability to >> re-trigger this scrub on-demand, in response to a memcpy_from_pmem() >> discovering an error, or after a SIGBUS is communicated to userspace. >> >> > - Is the filesystem supposed to intercept >> > reads/writes to bad blocks once it knows about >> > them? >> >> No, the driver handles that. > > So, -EIO will be returned to the filesystem on access? If -EIO, then > we'll have to check over the bad block list to determine if data > recovery operations are required, right? Perhaps we need a different > error here to tell the higher layers it's a specific type of error > (e.g. -EBADBLOCK)? That sounds reasonable. >> > - how is the filesystem supposed to communicate that >> > there is a bad block in a file back to userspace? >> >> -EIO on access. > > So no consideration for proactive "data loss has occurred at offset X > in file /mnt/path/to/file, attempting recovery" messages when the > error is first detected by the lowest layers? That sounds reasonable too, what does xfs report today for disk errors? >> > Or is userspace supposed to infer that there's a >> > bad block from EIO and so has to run FIEMAP to >> > determine if the error really was due to a bad >> > block? >> >> The information is there to potentially do forensics on why an I/O >> encountered an error, but there is no expectation that userspace >> follow up on each -EIO with a FIEMAP. > > Ok, so how is userspace driven error recovery supposed to work if it > can't differentiate the cause of an EIO error? If there's no > requirement for FIEMAP to report the bad blocks in a file that needs > recovery, then what is the app supposed to do with the EIO? Indeed, > what consideration has been given to ensuring the app knows aheadi > of time that the filesystem FIEMAP implementation will report bad > blocks if they exist? So, having a new flavor of FIEMAP that reports bad blocks in a file, or a new SEEK_BADBLOCK flag for lseek() was proposed for situations where an application could not simply take a list of badblocks from sysfs and do an intersection operation with typical FIEMAP results to see if a file was impacted. Indeed btrfs is an example where the per-block-device list is unusable for an application to do this FIEMAP intersection operation, but an application on a single-device-ext4 filesystem could make use of the list. > Of course, the filesystem has to know about the bad blocks to be > able to do any of this with FIEMAP.... Is there some complication about looking up the gendisk from the inode that I'm overlooking? Is there a different location we need to place the error list to make it easy for the fs to consume? >> > - what happens if there is no running application >> > that we can report the error to or will handle the >> > error (e.g. found error by a media scrub or during >> > boot)? >> >> Same as RAID today, if the array is in sync the bad block will get >> re-written during the scrub hopefully in advance of when an >> application might discover it. If no RAID is present then the only >> notification is an on-access error. > > More /assumptions/ that only device level RAID will be able to > recover.... Presently yes, in the future, no. > >> > - if the bad block is in filesystem free space, what should >> > the filesystem do with it? >> >> Nothing. When the free space becomes allocated we rely on the fact >> that the filesystem will first zero the blocks. That zeroing process >> will clear the media error. > > The incorrect /assumption/ here is that all allocations will do > block zeroing first. That's simply wrong. We do that for *user data* > in XFS and ext4, but we do not do it for metadata as they are not > accessed by DAX and, being transactionally protected, don't need > zeroing to prevent stale data exposure. Ah, great point. See, I knew if we kept talking, something productive would come out of this discussion, but I think we're still ok, see below. > Hence we have a problem here - the first write to such blocks may > be metadata writeback of some type and so the filesystem will see > EIO errors in metadata writes and they'll freak out. As long as the filesystem writes before it reads metadata we're mostly ok, because a block error is cleared on write, we don't return -EIO for them, however... > What > now - does this really mean that we'll have to add special IO > falback code for all internal IO paths to be able to clear pmem bad > block errors? ...we still have the case where a badblock can develop in metadata after writing it, we can't do anything about it unless some layer somewhere implements redundancy. > Oh, and just a thought: lots of people are pushing for selectable > FALLOC_FL_NO_HIDE_STALE behaviour which will skip zeroing > of data blocks on allocation. If this happens, it we also skip > the zeroing on allocation, so again there is no mechanism to clear > bad block status in this case..... Outside of rewriting the file, yes I was aware of this and it breaks the "delete to clear media errors" recovery model. My only answer right now is that we would need to document FALLOC_FL_NO_HIDE_STALE with notifications about the fact that we lose the side-effect of clearing latent media errors when block zeroing / trimming on reallocation is disabled. > [...] > >> > However, if I'm struggling to understand how I'm supposed to >> > connecct up the parts inside a filesytem, then expecting application >> > developers to be able to connect the dots in a sane manner is >> > bordering on fantasy.... >> >> Hopefully it is becoming clearer that we are not proposing anything >> radically different than what is present for error recovery today >> modulo thinking about the mechanisms to trap and recover a DAX read of >> a bad media area via a DM/MD implementation. > > There is a radical difference - there is a pervasive /assumption/ in > what is being proposed that filesystems are incapable of storing > redundant information that can be used for error recovery. That's not an assumption, it is a fact that DAX enabled filesystems don't account for data redundancy today outside of "let the storage layer do it". I'm all for making this situation better than it is. > The > current IO stack makes no such assumptions, even if it doesn't > provide any infrastructure for such functionality. Great, lets start to fill in that hole with some patches to return SIGBUS on a DAX fault hitting a bad block, and bypassing dax_do_io() for O_DIRECT writes so userspace I/O can clear errors instead of receiving -EIO for a write.
diff --git a/fs/block_dev.c b/fs/block_dev.c index c5837fa..d6113b9 100644 --- a/fs/block_dev.c +++ b/fs/block_dev.c @@ -166,13 +166,24 @@ blkdev_direct_IO(struct kiocb *iocb, struct iov_iter *iter, loff_t offset) { struct file *file = iocb->ki_filp; struct inode *inode = bdev_file_inode(file); + ssize_t ret, ret_saved = 0; - if (IS_DAX(inode)) - return dax_do_io(iocb, inode, iter, offset, blkdev_get_block, + if (IS_DAX(inode)) { + ret = dax_do_io(iocb, inode, iter, offset, blkdev_get_block, NULL, DIO_SKIP_DIO_COUNT); - return __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, offset, + if (ret == -EIO && (iov_iter_rw(iter) == WRITE)) + ret_saved = ret; + else + return ret; + } + + ret = __blockdev_direct_IO(iocb, inode, I_BDEV(inode), iter, offset, blkdev_get_block, NULL, NULL, DIO_SKIP_DIO_COUNT); + if (ret < 0 && ret_saved) + return ret_saved; + + return ret; } int __sync_blockdev(struct block_device *bdev, int wait) diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c index 824f249..64792c6 100644 --- a/fs/ext2/inode.c +++ b/fs/ext2/inode.c @@ -859,14 +859,22 @@ ext2_direct_IO(struct kiocb *iocb, struct iov_iter *iter, loff_t offset) struct address_space *mapping = file->f_mapping; struct inode *inode = mapping->host; size_t count = iov_iter_count(iter); - ssize_t ret; + ssize_t ret, ret_saved = 0; - if (IS_DAX(inode)) - ret = dax_do_io(iocb, inode, iter, offset, ext2_get_block, NULL, - DIO_LOCKING); - else - ret = blockdev_direct_IO(iocb, inode, iter, offset, - ext2_get_block); + if (IS_DAX(inode)) { + ret = dax_do_io(iocb, inode, iter, offset, ext2_get_block, + NULL, DIO_LOCKING | DIO_SKIP_HOLES); + if (ret == -EIO && iov_iter_rw(iter) == WRITE) + ret_saved = ret; + else + goto out; + } + + ret = blockdev_direct_IO(iocb, inode, iter, offset, ext2_get_block); + if (ret < 0 && ret_saved) + ret = ret_saved; + + out: if (ret < 0 && iov_iter_rw(iter) == WRITE) ext2_write_failed(mapping, offset + count); return ret; diff --git a/fs/ext4/indirect.c b/fs/ext4/indirect.c index 3027fa6..798f341 100644 --- a/fs/ext4/indirect.c +++ b/fs/ext4/indirect.c @@ -716,14 +716,22 @@ retry: NULL, NULL, 0); inode_dio_end(inode); } else { + ssize_t ret_saved = 0; + locked: - if (IS_DAX(inode)) + if (IS_DAX(inode)) { ret = dax_do_io(iocb, inode, iter, offset, ext4_dio_get_block, NULL, DIO_LOCKING); - else - ret = blockdev_direct_IO(iocb, inode, iter, offset, - ext4_dio_get_block); - + if (ret == -EIO && iov_iter_rw(iter) == WRITE) + ret_saved = ret; + else + goto skip_dio; + } + ret = blockdev_direct_IO(iocb, inode, iter, offset, + ext4_get_block); + if (ret < 0 && ret_saved) + ret = ret_saved; +skip_dio: if (unlikely(iov_iter_rw(iter) == WRITE && ret < 0)) { loff_t isize = i_size_read(inode); loff_t end = offset + count; diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index dab84a2..27f07c2 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -3341,7 +3341,7 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter, { struct file *file = iocb->ki_filp; struct inode *inode = file->f_mapping->host; - ssize_t ret; + ssize_t ret, ret_saved = 0; size_t count = iov_iter_count(iter); int overwrite = 0; get_block_t *get_block_func = NULL; @@ -3401,15 +3401,22 @@ static ssize_t ext4_ext_direct_IO(struct kiocb *iocb, struct iov_iter *iter, #ifdef CONFIG_EXT4_FS_ENCRYPTION BUG_ON(ext4_encrypted_inode(inode) && S_ISREG(inode->i_mode)); #endif - if (IS_DAX(inode)) + if (IS_DAX(inode)) { ret = dax_do_io(iocb, inode, iter, offset, get_block_func, ext4_end_io_dio, dio_flags); - else - ret = __blockdev_direct_IO(iocb, inode, - inode->i_sb->s_bdev, iter, offset, - get_block_func, - ext4_end_io_dio, NULL, dio_flags); + if (ret == -EIO && iov_iter_rw(iter) == WRITE) + ret_saved = ret; + else + goto skip_dio; + } + ret = __blockdev_direct_IO(iocb, inode, + inode->i_sb->s_bdev, iter, offset, + get_block_func, + ext4_end_io_dio, NULL, dio_flags); + if (ret < 0 && ret_saved) + ret = ret_saved; + skip_dio: if (ret > 0 && !overwrite && ext4_test_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN)) { int err; diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index d445a64..7cfcf86 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -1413,6 +1413,7 @@ xfs_vm_direct_IO( dio_iodone_t *endio = NULL; int flags = 0; struct block_device *bdev; + ssize_t ret, ret_saved = 0; if (iov_iter_rw(iter) == WRITE) { endio = xfs_end_io_direct_write; @@ -1420,13 +1421,22 @@ xfs_vm_direct_IO( } if (IS_DAX(inode)) { - return dax_do_io(iocb, inode, iter, offset, + ret = dax_do_io(iocb, inode, iter, offset, xfs_get_blocks_direct, endio, 0); + if (ret == -EIO && iov_iter_rw(iter) == WRITE) + ret_saved = ret; + else + return ret; } bdev = xfs_find_bdev_for_inode(inode); - return __blockdev_direct_IO(iocb, inode, bdev, iter, offset, + ret = __blockdev_direct_IO(iocb, inode, bdev, iter, offset, xfs_get_blocks_direct, endio, NULL, flags); + + if (ret < 0 && ret_saved) + ret = ret_saved; + + return ret; } /*
dax_do_io (called for read() or write() for a dax file system) may fail in the presence of bad blocks or media errors. Since we expect that a write should clear media errors on nvdimms, make dax_do_io fall back to the direct_IO path, which will send down a bio to the driver, which can then attempt to clear the error. Cc: Matthew Wilcox <matthew.r.wilcox@intel.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Ross Zwisler <ross.zwisler@linux.intel.com> Cc: Dave Chinner <david@fromorbit.com> Cc: Jan Kara <jack@suse.cz> Cc: Jens Axboe <axboe@fb.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Cc: Christoph Hellwig <hch@infradead.org> Signed-off-by: Vishal Verma <vishal.l.verma@intel.com> --- fs/block_dev.c | 17 ++++++++++++++--- fs/ext2/inode.c | 22 +++++++++++++++------- fs/ext4/indirect.c | 18 +++++++++++++----- fs/ext4/inode.c | 21 ++++++++++++++------- fs/xfs/xfs_aops.c | 14 ++++++++++++-- 5 files changed, 68 insertions(+), 24 deletions(-)