| Message ID | 20190218100433.20048-1-chandan@linux.ibm.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Consolidate Post read processing code | expand |
On Monday, February 18, 2019 3:34:23 PM IST Chandan Rajendra wrote: > This patchset moves the "post read processing" code into a file of its > own and gets the generic do_mpage_readpge() to make use of the > functionality provided. With these changes in place, the patchset > changes Ext4 to use mpage_readpage[s] instead of its own custom > ext4_readpages() function. This is done to reduce duplicity of code > across filesystems. Based on the reviews provided for this patchset, I > will change F2FS to use mpage_readpage[s] and post the next version of > this patchset to linux-fsdevel mailing list. > > The patchset also includes patches from previous postings i.e. > patches to replace per-filesystem encryption config options with a > single config option that affects all filesystems making use of > fscrypt code. The patchset is based on fsverity_2019-01-30 branch of Eric's git tree. It can also be obtained from "https://github.com/chandanr/linux.git use-mpage-readpage-rfc". > > Chandan Rajendra (10): > ext4: use IS_ENCRYPTED() to check encryption status > f2fs: use IS_ENCRYPTED() to check encryption status > fscrypt: remove filesystem specific build config option > Consolidate "post read processing" into a new file > fsverity: Add call back to decide if verity check has to be performed > Introduce REQ_POST_READ_PROC bio flag > fsverity: Add call back to determine readpage limit > fsverity: Add call back to verify file holes > fs/mpage.c: Integrate post read processing > ext4: Wire up ext4_readpage[s] to use mpage_readpage[s] > > Documentation/filesystems/fscrypt.rst | 4 +- > arch/mips/configs/generic_defconfig | 2 +- > arch/nds32/configs/defconfig | 2 +- > arch/s390/configs/debug_defconfig | 2 +- > arch/s390/configs/performance_defconfig | 2 +- > fs/Makefile | 3 +- > fs/crypto/Kconfig | 5 +- > fs/crypto/bio.c | 21 +- > fs/crypto/crypto.c | 1 + > fs/crypto/fscrypt_private.h | 4 +- > fs/ext4/Kconfig | 15 - > fs/ext4/Makefile | 2 +- > fs/ext4/dir.c | 10 +- > fs/ext4/ext4.h | 14 +- > fs/ext4/ext4_jbd2.h | 2 +- > fs/ext4/extents.c | 4 +- > fs/ext4/ialloc.c | 2 +- > fs/ext4/inode.c | 29 +- > fs/ext4/ioctl.c | 4 +- > fs/ext4/move_extent.c | 3 +- > fs/ext4/namei.c | 18 +- > fs/ext4/page-io.c | 9 +- > fs/ext4/readpage.c | 444 ------------------------ > fs/ext4/super.c | 43 ++- > fs/ext4/sysfs.c | 4 +- > fs/f2fs/Kconfig | 12 +- > fs/f2fs/data.c | 4 +- > fs/f2fs/dir.c | 10 +- > fs/f2fs/f2fs.h | 14 +- > fs/f2fs/file.c | 10 +- > fs/f2fs/inode.c | 4 +- > fs/f2fs/namei.c | 6 +- > fs/f2fs/super.c | 8 +- > fs/f2fs/sysfs.c | 4 +- > fs/mpage.c | 77 +++- > fs/post_read_process.c | 128 +++++++ > fs/ubifs/Kconfig | 12 +- > fs/ubifs/Makefile | 2 +- > fs/ubifs/ioctl.c | 4 +- > fs/ubifs/sb.c | 2 +- > fs/ubifs/super.c | 2 +- > fs/ubifs/ubifs.h | 5 +- > fs/verity/verify.c | 12 + > include/linux/blk_types.h | 2 + > include/linux/fs.h | 4 +- > include/linux/fscrypt.h | 405 ++++++++++++++++++++- > include/linux/fscrypt_notsupp.h | 231 ------------ > include/linux/fscrypt_supp.h | 204 ----------- > include/linux/fsverity.h | 3 + > include/linux/post_read_process.h | 21 ++ > 50 files changed, 757 insertions(+), 1078 deletions(-) > delete mode 100644 fs/ext4/readpage.c > create mode 100644 fs/post_read_process.c > delete mode 100644 include/linux/fscrypt_notsupp.h > delete mode 100644 include/linux/fscrypt_supp.h > create mode 100644 include/linux/post_read_process.h > >
Hi Chandan, On Mon, Feb 18, 2019 at 03:34:23PM +0530, Chandan Rajendra wrote: > This patchset moves the "post read processing" code into a file of its > own and gets the generic do_mpage_readpge() to make use of the > functionality provided. With these changes in place, the patchset > changes Ext4 to use mpage_readpage[s] instead of its own custom > ext4_readpages() function. This is done to reduce duplicity of code > across filesystems. Based on the reviews provided for this patchset, I > will change F2FS to use mpage_readpage[s] and post the next version of > this patchset to linux-fsdevel mailing list. > > The patchset also includes patches from previous postings i.e. > patches to replace per-filesystem encryption config options with a > single config option that affects all filesystems making use of > fscrypt code. > > Chandan Rajendra (10): > ext4: use IS_ENCRYPTED() to check encryption status > f2fs: use IS_ENCRYPTED() to check encryption status > fscrypt: remove filesystem specific build config option > Consolidate "post read processing" into a new file > fsverity: Add call back to decide if verity check has to be performed > Introduce REQ_POST_READ_PROC bio flag > fsverity: Add call back to determine readpage limit > fsverity: Add call back to verify file holes > fs/mpage.c: Integrate post read processing > ext4: Wire up ext4_readpage[s] to use mpage_readpage[s] > Thanks for working on this! This will also make it much easier to support block_size != PAGE_SIZE in ext4 encryption, right? I think this is the best path forward, but I'll take a closer look at your new patches. FYI regarding practical matters, merging fs-verity was delayed due to disagreement about the API. See https://lwn.net/Articles/775872/. We don't have to wait for fs-verity for your initial fscrypt changes, though: ext4: use IS_ENCRYPTED() to check encryption status f2fs: use IS_ENCRYPTED() to check encryption status fscrypt: remove filesystem specific build config option So, a couple weeks ago Ted and I already queued those three patches in fscrypt.git (https://git.kernel.org/pub/scm/linux/kernel/git/tytso/fscrypt.git branch "master", though we plan to change the repo soon) for the upcoming merge window, based on upstream rather than fs-verity. Are you fine with that? I also suggest adding linux-fsdevel to the Cc given the fs/*.c changes. Thanks! - Eric
On Mon, Feb 18, 2019 at 03:34:23PM +0530, Chandan Rajendra wrote: > Based on the reviews provided for this patchset, I > will change F2FS to use mpage_readpage[s] and post the next version of > this patchset to linux-fsdevel mailing list. Is that really possible? F2FS has other functionality in its ->readpages(), such as support for filesystems that use multiple block devices. - Eric
On Wednesday, February 20, 2019 2:47:16 AM IST Eric Biggers wrote: > Hi Chandan, > > On Mon, Feb 18, 2019 at 03:34:23PM +0530, Chandan Rajendra wrote: > > This patchset moves the "post read processing" code into a file of its > > own and gets the generic do_mpage_readpge() to make use of the > > functionality provided. With these changes in place, the patchset > > changes Ext4 to use mpage_readpage[s] instead of its own custom > > ext4_readpages() function. This is done to reduce duplicity of code > > across filesystems. Based on the reviews provided for this patchset, I > > will change F2FS to use mpage_readpage[s] and post the next version of > > this patchset to linux-fsdevel mailing list. > > > > The patchset also includes patches from previous postings i.e. > > patches to replace per-filesystem encryption config options with a > > single config option that affects all filesystems making use of > > fscrypt code. > > > > Chandan Rajendra (10): > > ext4: use IS_ENCRYPTED() to check encryption status > > f2fs: use IS_ENCRYPTED() to check encryption status > > fscrypt: remove filesystem specific build config option > > Consolidate "post read processing" into a new file > > fsverity: Add call back to decide if verity check has to be performed > > Introduce REQ_POST_READ_PROC bio flag > > fsverity: Add call back to determine readpage limit > > fsverity: Add call back to verify file holes > > fs/mpage.c: Integrate post read processing > > ext4: Wire up ext4_readpage[s] to use mpage_readpage[s] > > > > Thanks for working on this! This will also make it much easier to support > block_size != PAGE_SIZE in ext4 encryption, right? I think this is the best > path forward, but I'll take a closer look at your new patches. > > FYI regarding practical matters, merging fs-verity was delayed due to > disagreement about the API. See https://lwn.net/Articles/775872/. > > We don't have to wait for fs-verity for your initial fscrypt changes, though: > > ext4: use IS_ENCRYPTED() to check encryption status > f2fs: use IS_ENCRYPTED() to check encryption status > fscrypt: remove filesystem specific build config option > > So, a couple weeks ago Ted and I already queued those three patches in > fscrypt.git (https://git.kernel.org/pub/scm/linux/kernel/git/tytso/fscrypt.git > branch "master", though we plan to change the repo soon) for the upcoming merge > window, based on upstream rather than fs-verity. Are you fine with that? Yes, the changes looks good. Thanks for queueing them up. > > I also suggest adding linux-fsdevel to the Cc given the fs/*.c changes. Yes, I will do that. > > Thanks! > > - Eric > >
On Wednesday, February 20, 2019 6:11:07 AM IST Eric Biggers wrote: > On Mon, Feb 18, 2019 at 03:34:23PM +0530, Chandan Rajendra wrote: > > Based on the reviews provided for this patchset, I > > will change F2FS to use mpage_readpage[s] and post the next version of > > this patchset to linux-fsdevel mailing list. > > Is that really possible? F2FS has other functionality in its ->readpages(), > such as support for filesystems that use multiple block devices. > Hmm. I didn't know that. Thanks for pointing that out. I will read up that code and if it isn't really possible to get F2FS to use mpage_readpage[s](), then I would limit the patchset to Ext4.
On Thu, Feb 21, 2019 at 07:02:03PM +0530, Chandan Rajendra wrote: > On Wednesday, February 20, 2019 6:11:07 AM IST Eric Biggers wrote: > > On Mon, Feb 18, 2019 at 03:34:23PM +0530, Chandan Rajendra wrote: > > > Based on the reviews provided for this patchset, I > > > will change F2FS to use mpage_readpage[s] and post the next version of > > > this patchset to linux-fsdevel mailing list. > > > > Is that really possible? F2FS has other functionality in its ->readpages(), > > such as support for filesystems that use multiple block devices. > > > > Hmm. I didn't know that. Thanks for pointing that out. I will read up that > code and if it isn't really possible to get F2FS to use mpage_readpage[s](), > then I would limit the patchset to Ext4. > But we can make F2FS use the shared bio_post_read functions even if it can't use mpage_readpage[s](), right? - Eric
On Friday, February 22, 2019 12:08:28 AM IST Eric Biggers wrote: > On Thu, Feb 21, 2019 at 07:02:03PM +0530, Chandan Rajendra wrote: > > On Wednesday, February 20, 2019 6:11:07 AM IST Eric Biggers wrote: > > > On Mon, Feb 18, 2019 at 03:34:23PM +0530, Chandan Rajendra wrote: > > > > Based on the reviews provided for this patchset, I > > > > will change F2FS to use mpage_readpage[s] and post the next version of > > > > this patchset to linux-fsdevel mailing list. > > > > > > Is that really possible? F2FS has other functionality in its ->readpages(), > > > such as support for filesystems that use multiple block devices. > > > > > > > Hmm. I didn't know that. Thanks for pointing that out. I will read up that > > code and if it isn't really possible to get F2FS to use mpage_readpage[s](), > > then I would limit the patchset to Ext4. > > > > But we can make F2FS use the shared bio_post_read functions even if it can't use > mpage_readpage[s](), right? > Yes, that should be possible. I will include those changes in the next version of the patchset.