| Message ID | 20241107235614.3637221-7-joannelkoong@gmail.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | fuse: remove temp page copies in writeback | expand |
Hi, Joanne, Thanks for the continuing work! On 11/8/24 7:56 AM, Joanne Koong wrote: > Currently, we allocate and copy data to a temporary folio when > handling writeback in order to mitigate the following deadlock scenario > that may arise if reclaim waits on writeback to complete: > * single-threaded FUSE server is in the middle of handling a request > that needs a memory allocation > * memory allocation triggers direct reclaim > * direct reclaim waits on a folio under writeback > * the FUSE server can't write back the folio since it's stuck in > direct reclaim > > To work around this, we allocate a temporary folio and copy over the > original folio to the temporary folio so that writeback can be > immediately cleared on the original folio. This additionally requires us > to maintain an internal rb tree to keep track of writeback state on the > temporary folios. > > A recent change prevents reclaim logic from waiting on writeback for > folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > will prevent FUSE folios from running into the reclaim deadlock described > above) and removes the temporary folio + extra copying and the internal > rb tree. > > fio benchmarks -- > (using averages observed from 10 runs, throwing away outliers) > > Setup: > sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > > fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > > bs = 1k 4k 1M > Before 351 MiB/s 1818 MiB/s 1851 MiB/s > After 341 MiB/s 2246 MiB/s 2685 MiB/s > % diff -3% 23% 45% > > Signed-off-by: Joanne Koong <joannelkoong@gmail.com> > @@ -1622,7 +1543,7 @@ ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter, > return res; > } > } > - if (!cuse && fuse_range_is_writeback(inode, idx_from, idx_to)) { > + if (!cuse && filemap_range_has_writeback(mapping, pos, (pos + count - 1))) { filemap_range_has_writeback() is not equivalent to fuse_range_is_writeback(), as it will return true as long as there's any locked or dirty page? I can't find an equivalent helper function at hand though. > @@ -3423,7 +3143,6 @@ void fuse_init_file_inode(struct inode *inode, unsigned int flags) > fi->iocachectr = 0; > init_waitqueue_head(&fi->page_waitq); > init_waitqueue_head(&fi->direct_io_waitq); > - fi->writepages = RB_ROOT; It seems that 'struct rb_root writepages' is not removed from fuse_inode structure. Besides, I also looked through the former 5 patches and can't find any obvious errors at the very first glance. Hopefully the MM guys could offer more professional reviews.
On Fri, Nov 8, 2024 at 12:49 AM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: > > Hi, Joanne, > > Thanks for the continuing work! > > On 11/8/24 7:56 AM, Joanne Koong wrote: > > Currently, we allocate and copy data to a temporary folio when > > handling writeback in order to mitigate the following deadlock scenario > > that may arise if reclaim waits on writeback to complete: > > * single-threaded FUSE server is in the middle of handling a request > > that needs a memory allocation > > * memory allocation triggers direct reclaim > > * direct reclaim waits on a folio under writeback > > * the FUSE server can't write back the folio since it's stuck in > > direct reclaim > > > > To work around this, we allocate a temporary folio and copy over the > > original folio to the temporary folio so that writeback can be > > immediately cleared on the original folio. This additionally requires us > > to maintain an internal rb tree to keep track of writeback state on the > > temporary folios. > > > > A recent change prevents reclaim logic from waiting on writeback for > > folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > > This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > > will prevent FUSE folios from running into the reclaim deadlock described > > above) and removes the temporary folio + extra copying and the internal > > rb tree. > > > > fio benchmarks -- > > (using averages observed from 10 runs, throwing away outliers) > > > > Setup: > > sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > > ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > > > > fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > > --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > > > > bs = 1k 4k 1M > > Before 351 MiB/s 1818 MiB/s 1851 MiB/s > > After 341 MiB/s 2246 MiB/s 2685 MiB/s > > % diff -3% 23% 45% > > > > Signed-off-by: Joanne Koong <joannelkoong@gmail.com> > > > > @@ -1622,7 +1543,7 @@ ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter, > > return res; > > } > > } > > - if (!cuse && fuse_range_is_writeback(inode, idx_from, idx_to)) { > > + if (!cuse && filemap_range_has_writeback(mapping, pos, (pos + count - 1))) { > > filemap_range_has_writeback() is not equivalent to > fuse_range_is_writeback(), as it will return true as long as there's any > locked or dirty page? I can't find an equivalent helper function at > hand though. > Hi Jingbo, I couldn't find an equivalent helper function either. My guess is that filemap_range_has_writeback() returns true if the page is locked because it doesn't have a way of determining if the page is dirty or not (it seems like if a page is locked, then we can't read the writeback bit on it) so it errs on the side of assuming yes. For this case, it seems okay to me to use filemap_range_has_writeback() because if we get back a false positive (eg filemap_range_has_writeback() returns true when it's actually false), the only cost is the overhead of an additional fuse_sync_writes() call but fuse_sync_writes() will return immediately from the wait(). > > > > @@ -3423,7 +3143,6 @@ void fuse_init_file_inode(struct inode *inode, unsigned int flags) > > fi->iocachectr = 0; > > init_waitqueue_head(&fi->page_waitq); > > init_waitqueue_head(&fi->direct_io_waitq); > > - fi->writepages = RB_ROOT; > > It seems that 'struct rb_root writepages' is not removed from fuse_inode > structure. > Nice catch! I'll remove this from the fuse_inode struct in v5. > > Besides, I also looked through the former 5 patches and can't find any > obvious errors at the very first glance. Hopefully the MM guys could > offer more professional reviews. > Thanks for looking through this code in this version and the past versions of this patchset too. It's much appreciated! > -- > Thanks, > Jingbo
Hi, Joanne and Miklos, On 11/8/24 7:56 AM, Joanne Koong wrote: > Currently, we allocate and copy data to a temporary folio when > handling writeback in order to mitigate the following deadlock scenario > that may arise if reclaim waits on writeback to complete: > * single-threaded FUSE server is in the middle of handling a request > that needs a memory allocation > * memory allocation triggers direct reclaim > * direct reclaim waits on a folio under writeback > * the FUSE server can't write back the folio since it's stuck in > direct reclaim > > To work around this, we allocate a temporary folio and copy over the > original folio to the temporary folio so that writeback can be > immediately cleared on the original folio. This additionally requires us > to maintain an internal rb tree to keep track of writeback state on the > temporary folios. > > A recent change prevents reclaim logic from waiting on writeback for > folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > will prevent FUSE folios from running into the reclaim deadlock described > above) and removes the temporary folio + extra copying and the internal > rb tree. > > fio benchmarks -- > (using averages observed from 10 runs, throwing away outliers) > > Setup: > sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > > fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > > bs = 1k 4k 1M > Before 351 MiB/s 1818 MiB/s 1851 MiB/s > After 341 MiB/s 2246 MiB/s 2685 MiB/s > % diff -3% 23% 45% > > Signed-off-by: Joanne Koong <joannelkoong@gmail.com> IIUC this patch seems to break commit 8b284dc47291daf72fe300e1138a2e7ed56f38ab ("fuse: writepages: handle same page rewrites"). > - /* > - * Being under writeback is unlikely but possible. For example direct > - * read to an mmaped fuse file will set the page dirty twice; once when > - * the pages are faulted with get_user_pages(), and then after the read > - * completed. > - */ In short, the target scenario is like: ``` # open a fuse file and mmap fd1 = open("fuse-file-path", ...) uaddr = mmap(fd1, ...) # DIRECT read to the mmaped fuse file fd2 = open("ext4-file-path", O_DIRECT, ...) read(fd2, uaddr, ...) # get_user_pages() of uaddr, and triggers faultin # a_ops->dirty_folio() <--- mark PG_dirty # when DIRECT IO completed: # a_ops->dirty_folio() <--- mark PG_dirty ``` The auxiliary write request list was introduced to fix this. I'm not sure if there's an alternative other than the auxiliary list to fix it, e.g. calling folio_wait_writeback() in a_ops->dirty_folio() so that the same folio won't get dirtied when the writeback has not completed yet?
On Mon, Nov 11, 2024 at 12:32 AM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: > > Hi, Joanne and Miklos, > > On 11/8/24 7:56 AM, Joanne Koong wrote: > > Currently, we allocate and copy data to a temporary folio when > > handling writeback in order to mitigate the following deadlock scenario > > that may arise if reclaim waits on writeback to complete: > > * single-threaded FUSE server is in the middle of handling a request > > that needs a memory allocation > > * memory allocation triggers direct reclaim > > * direct reclaim waits on a folio under writeback > > * the FUSE server can't write back the folio since it's stuck in > > direct reclaim > > > > To work around this, we allocate a temporary folio and copy over the > > original folio to the temporary folio so that writeback can be > > immediately cleared on the original folio. This additionally requires us > > to maintain an internal rb tree to keep track of writeback state on the > > temporary folios. > > > > A recent change prevents reclaim logic from waiting on writeback for > > folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > > This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > > will prevent FUSE folios from running into the reclaim deadlock described > > above) and removes the temporary folio + extra copying and the internal > > rb tree. > > > > fio benchmarks -- > > (using averages observed from 10 runs, throwing away outliers) > > > > Setup: > > sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > > ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > > > > fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > > --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > > > > bs = 1k 4k 1M > > Before 351 MiB/s 1818 MiB/s 1851 MiB/s > > After 341 MiB/s 2246 MiB/s 2685 MiB/s > > % diff -3% 23% 45% > > > > Signed-off-by: Joanne Koong <joannelkoong@gmail.com> > > > IIUC this patch seems to break commit > 8b284dc47291daf72fe300e1138a2e7ed56f38ab ("fuse: writepages: handle same > page rewrites"). > Interesting! My understanding was that we only needed that commit because we were clearing writeback on the original folio before writeback had actually finished. Now that folio writeback state is accounted for normally (eg through writeback being set/cleared on the original folio), does the folio_wait_writeback() call we do in fuse_page_mkwrite() not mitigate this? > > - /* > > - * Being under writeback is unlikely but possible. For example direct > > - * read to an mmaped fuse file will set the page dirty twice; once when > > - * the pages are faulted with get_user_pages(), and then after the read > > - * completed. > > - */ > > In short, the target scenario is like: > > ``` > # open a fuse file and mmap > fd1 = open("fuse-file-path", ...) > uaddr = mmap(fd1, ...) > > # DIRECT read to the mmaped fuse file > fd2 = open("ext4-file-path", O_DIRECT, ...) > read(fd2, uaddr, ...) > # get_user_pages() of uaddr, and triggers faultin > # a_ops->dirty_folio() <--- mark PG_dirty > > # when DIRECT IO completed: > # a_ops->dirty_folio() <--- mark PG_dirty If you have the direct io function call stack at hand, could you point me to the function where the direct io completion marks this folio as dirty? > ``` > > The auxiliary write request list was introduced to fix this. > > I'm not sure if there's an alternative other than the auxiliary list to > fix it, e.g. calling folio_wait_writeback() in a_ops->dirty_folio() so > that the same folio won't get dirtied when the writeback has not > completed yet? > I'm curious how other filesystems solve for this - this seems like a generic situation other filesystems would run into as well. Thanks, Joanne > > > -- > Thanks, > Jingbo
On 11/12/24 5:30 AM, Joanne Koong wrote: > On Mon, Nov 11, 2024 at 12:32 AM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: >> >> Hi, Joanne and Miklos, >> >> On 11/8/24 7:56 AM, Joanne Koong wrote: >>> Currently, we allocate and copy data to a temporary folio when >>> handling writeback in order to mitigate the following deadlock scenario >>> that may arise if reclaim waits on writeback to complete: >>> * single-threaded FUSE server is in the middle of handling a request >>> that needs a memory allocation >>> * memory allocation triggers direct reclaim >>> * direct reclaim waits on a folio under writeback >>> * the FUSE server can't write back the folio since it's stuck in >>> direct reclaim >>> >>> To work around this, we allocate a temporary folio and copy over the >>> original folio to the temporary folio so that writeback can be >>> immediately cleared on the original folio. This additionally requires us >>> to maintain an internal rb tree to keep track of writeback state on the >>> temporary folios. >>> >>> A recent change prevents reclaim logic from waiting on writeback for >>> folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. >>> This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which >>> will prevent FUSE folios from running into the reclaim deadlock described >>> above) and removes the temporary folio + extra copying and the internal >>> rb tree. >>> >>> fio benchmarks -- >>> (using averages observed from 10 runs, throwing away outliers) >>> >>> Setup: >>> sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount >>> ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount >>> >>> fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G >>> --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount >>> >>> bs = 1k 4k 1M >>> Before 351 MiB/s 1818 MiB/s 1851 MiB/s >>> After 341 MiB/s 2246 MiB/s 2685 MiB/s >>> % diff -3% 23% 45% >>> >>> Signed-off-by: Joanne Koong <joannelkoong@gmail.com> >> >> >> IIUC this patch seems to break commit >> 8b284dc47291daf72fe300e1138a2e7ed56f38ab ("fuse: writepages: handle same >> page rewrites"). >> > > Interesting! My understanding was that we only needed that commit > because we were clearing writeback on the original folio before > writeback had actually finished. > > Now that folio writeback state is accounted for normally (eg through > writeback being set/cleared on the original folio), does the > folio_wait_writeback() call we do in fuse_page_mkwrite() not mitigate > this? Yes, after inspecting the writeback logic more, it seems that the second writeback won't be initiated if the first one has not completed yet, see ``` a_ops->writepages write_cache_pages writeback_iter writeback_get_folio folio_prepare_writeback if folio_test_writeback(folio): folio_wait_writeback(folio) ``` and thus it won't be an issue to remove the auxiliary list ;) > >>> - /* >>> - * Being under writeback is unlikely but possible. For example direct >>> - * read to an mmaped fuse file will set the page dirty twice; once when >>> - * the pages are faulted with get_user_pages(), and then after the read >>> - * completed. >>> - */ >> >> In short, the target scenario is like: >> >> ``` >> # open a fuse file and mmap >> fd1 = open("fuse-file-path", ...) >> uaddr = mmap(fd1, ...) >> >> # DIRECT read to the mmaped fuse file >> fd2 = open("ext4-file-path", O_DIRECT, ...) >> read(fd2, uaddr, ...) >> # get_user_pages() of uaddr, and triggers faultin >> # a_ops->dirty_folio() <--- mark PG_dirty >> >> # when DIRECT IO completed: >> # a_ops->dirty_folio() <--- mark PG_dirty > > If you have the direct io function call stack at hand, could you point > me to the function where the direct io completion marks this folio as > dirty? FYI The full call stack is like: ``` # DIRECT read(2) to the mmaped fuse file read(fd2, uaddr1, ...) f_ops->read_iter() (iomap-based ) iomap_dio_rw # for READ && user_backed_iter(iter): dio->flags |= IOMAP_DIO_DIRTY iomap_dio_iter iomap_dio_bio_iter # add user or kernel pages to a bio bio_iov_iter_get_pages ... pin_user_pages_fast(..., FOLL_WRITE, ...) # find corresponding vma of dest buffer (fuse page cache) # search page table (pet) to find corresponding page # if not fault yet, trigger explicit faultin: faultin_page(..., FOLL_WRITE, ...) handle_mm_fault(..., FAULT_FLAG_WRITE) handle_pte_fault do_wp_page (vma->vm_flags & VM_SHARED) wp_page_shared ... fault_dirty_shared_page folio_mark_dirty a_ops->dirty_folio(), i.e., filemap_dirty_folio() # set PG_dirty folio_test_set_dirty(folio) # set PAGECACHE_TAG_DIRTY __folio_mark_dirty # if dio->flags & IOMAP_DIO_DIRTY: bio_set_pages_dirty (for each dest page) folio_mark_dirty a_ops->dirty_folio(), i.e., filemap_dirty_folio() # set PG_dirty folio_test_set_dirty(folio) # set PAGECACHE_TAG_DIRTY __folio_mark_dirty ``` > >> ``` >> >> The auxiliary write request list was introduced to fix this. >> >> I'm not sure if there's an alternative other than the auxiliary list to >> fix it, e.g. calling folio_wait_writeback() in a_ops->dirty_folio() so >> that the same folio won't get dirtied when the writeback has not >> completed yet? >> > > I'm curious how other filesystems solve for this - this seems like a > generic situation other filesystems would run into as well. > As mentioned above, the writeback path will prevent the duplicate writeback request on the same page when the first writeback IO has not completed yet. Sorry for the noise...
Hi Joanne, On 11/8/24 7:56 AM, Joanne Koong wrote: > Currently, we allocate and copy data to a temporary folio when > handling writeback in order to mitigate the following deadlock scenario > that may arise if reclaim waits on writeback to complete: > * single-threaded FUSE server is in the middle of handling a request > that needs a memory allocation > * memory allocation triggers direct reclaim > * direct reclaim waits on a folio under writeback > * the FUSE server can't write back the folio since it's stuck in > direct reclaim > > To work around this, we allocate a temporary folio and copy over the > original folio to the temporary folio so that writeback can be > immediately cleared on the original folio. This additionally requires us > to maintain an internal rb tree to keep track of writeback state on the > temporary folios. > > A recent change prevents reclaim logic from waiting on writeback for > folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > will prevent FUSE folios from running into the reclaim deadlock described > above) and removes the temporary folio + extra copying and the internal > rb tree. > > fio benchmarks -- > (using averages observed from 10 runs, throwing away outliers) > > Setup: > sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > > fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > > bs = 1k 4k 1M > Before 351 MiB/s 1818 MiB/s 1851 MiB/s > After 341 MiB/s 2246 MiB/s 2685 MiB/s > % diff -3% 23% 45% > > Signed-off-by: Joanne Koong <joannelkoong@gmail.com> I think there are some places checking or waiting for writeback could be reconsidered if they are still needed or not after we dropping the temp page design. As they are inherited from the original implementation, at least they are harmless. I think they could be remained in this patch, and could be cleaned up later if really needed. > @@ -891,7 +813,7 @@ static int fuse_do_readfolio(struct file *file, struct folio *folio) > * have writeback that extends beyond the lifetime of the folio. So > * make sure we read a properly synced folio. > */ > - fuse_wait_on_folio_writeback(inode, folio); > + folio_wait_writeback(folio); I doubt if wait-on-writeback is needed here, as now page cache won't be freed until the writeback IO completes. The routine attempts to free page cache, e.g. invalidate_inode_pages2() (generally called by distributed filesystems when the file content has been modified from remote) or truncate_inode_pages() (called from truncate(2) or inode eviction) will wait for writeback completion (if any) before freeing page. Thus I don't think there's any possibility that .read_folio() or .readahead() will be called when the writeback has not completed. > @@ -1172,7 +1093,7 @@ static ssize_t fuse_send_write_pages(struct fuse_io_args *ia, > int err; > > for (i = 0; i < ap->num_folios; i++) > - fuse_wait_on_folio_writeback(inode, ap->folios[i]); > + folio_wait_writeback(ap->folios[i]); Ditto. > static void fuse_writepage_args_page_fill(struct fuse_writepage_args *wpa, struct folio *folio, > - struct folio *tmp_folio, uint32_t folio_index) > + uint32_t folio_index) > { > struct inode *inode = folio->mapping->host; > struct fuse_args_pages *ap = &wpa->ia.ap; > > - folio_copy(tmp_folio, folio); > - > - ap->folios[folio_index] = tmp_folio; > + folio_get(folio); I still think this folio_get() here is harmless but redundant. Ditto page cache won't be freed before writeback completes. Besides, other .writepages() implementaions e.g. iomap_writepages() also doen't get the refcount when constructing the writeback IO. > @@ -2481,7 +2200,7 @@ static int fuse_write_begin(struct file *file, struct address_space *mapping, > if (IS_ERR(folio)) > goto error; > > - fuse_wait_on_page_writeback(mapping->host, folio->index); > + folio_wait_writeback(folio); I also doubt if wait_on_writeback() is needed here, as now there won't be duplicate writeback IOs for the same page. > @@ -2545,13 +2264,11 @@ static int fuse_launder_folio(struct folio *folio) > { > int err = 0; > if (folio_clear_dirty_for_io(folio)) { > - struct inode *inode = folio->mapping->host; > - > /* Serialize with pending writeback for the same page */ > - fuse_wait_on_page_writeback(inode, folio->index); > + folio_wait_writeback(folio); I think folio_wait_writeback() is unneeded after dropping the temp page copying. This is introduced in commit 3993382bb3198cc5e263c3519418e716bd57b056 ("fuse: launder page should wait for page writeback") since .launder_page() could be called when the previous writeback of the same page has not completed yet. Since now we won't clear PG_writeback until the writeback completes, .launder_page() won't be called on the same page when the corresponding writeback IO is still inflight.
On Mon, Nov 11, 2024 at 6:31 PM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: > > > > On 11/12/24 5:30 AM, Joanne Koong wrote: > > On Mon, Nov 11, 2024 at 12:32 AM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: > >> > >> Hi, Joanne and Miklos, > >> > >> On 11/8/24 7:56 AM, Joanne Koong wrote: > >>> Currently, we allocate and copy data to a temporary folio when > >>> handling writeback in order to mitigate the following deadlock scenario > >>> that may arise if reclaim waits on writeback to complete: > >>> * single-threaded FUSE server is in the middle of handling a request > >>> that needs a memory allocation > >>> * memory allocation triggers direct reclaim > >>> * direct reclaim waits on a folio under writeback > >>> * the FUSE server can't write back the folio since it's stuck in > >>> direct reclaim > >>> > >>> To work around this, we allocate a temporary folio and copy over the > >>> original folio to the temporary folio so that writeback can be > >>> immediately cleared on the original folio. This additionally requires us > >>> to maintain an internal rb tree to keep track of writeback state on the > >>> temporary folios. > >>> > >>> A recent change prevents reclaim logic from waiting on writeback for > >>> folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > >>> This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > >>> will prevent FUSE folios from running into the reclaim deadlock described > >>> above) and removes the temporary folio + extra copying and the internal > >>> rb tree. > >>> > >>> fio benchmarks -- > >>> (using averages observed from 10 runs, throwing away outliers) > >>> > >>> Setup: > >>> sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > >>> ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > >>> > >>> fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > >>> --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > >>> > >>> bs = 1k 4k 1M > >>> Before 351 MiB/s 1818 MiB/s 1851 MiB/s > >>> After 341 MiB/s 2246 MiB/s 2685 MiB/s > >>> % diff -3% 23% 45% > >>> > >>> Signed-off-by: Joanne Koong <joannelkoong@gmail.com> > >> > >> > >> IIUC this patch seems to break commit > >> 8b284dc47291daf72fe300e1138a2e7ed56f38ab ("fuse: writepages: handle same > >> page rewrites"). > >> > > > > Interesting! My understanding was that we only needed that commit > > because we were clearing writeback on the original folio before > > writeback had actually finished. > > > > Now that folio writeback state is accounted for normally (eg through > > writeback being set/cleared on the original folio), does the > > folio_wait_writeback() call we do in fuse_page_mkwrite() not mitigate > > this? > > Yes, after inspecting the writeback logic more, it seems that the second > writeback won't be initiated if the first one has not completed yet, see > > ``` > a_ops->writepages > write_cache_pages > writeback_iter > writeback_get_folio > folio_prepare_writeback > if folio_test_writeback(folio): > folio_wait_writeback(folio) > ``` > > and thus it won't be an issue to remove the auxiliary list ;) > Awesome, thanks for double-checking! > > > >>> - /* > >>> - * Being under writeback is unlikely but possible. For example direct > >>> - * read to an mmaped fuse file will set the page dirty twice; once when > >>> - * the pages are faulted with get_user_pages(), and then after the read > >>> - * completed. > >>> - */ > >> > >> In short, the target scenario is like: > >> > >> ``` > >> # open a fuse file and mmap > >> fd1 = open("fuse-file-path", ...) > >> uaddr = mmap(fd1, ...) > >> > >> # DIRECT read to the mmaped fuse file > >> fd2 = open("ext4-file-path", O_DIRECT, ...) > >> read(fd2, uaddr, ...) > >> # get_user_pages() of uaddr, and triggers faultin > >> # a_ops->dirty_folio() <--- mark PG_dirty > >> > >> # when DIRECT IO completed: > >> # a_ops->dirty_folio() <--- mark PG_dirty > > > > If you have the direct io function call stack at hand, could you point > > me to the function where the direct io completion marks this folio as > > dirty? > > > FYI The full call stack is like: > > ``` > # DIRECT read(2) to the mmaped fuse file > read(fd2, uaddr1, ...) > f_ops->read_iter() > (iomap-based ) iomap_dio_rw > # for READ && user_backed_iter(iter): > dio->flags |= IOMAP_DIO_DIRTY > iomap_dio_iter > iomap_dio_bio_iter > # add user or kernel pages to a bio > bio_iov_iter_get_pages > ... > pin_user_pages_fast(..., FOLL_WRITE, ...) > # find corresponding vma of dest buffer (fuse page cache) > # search page table (pet) to find corresponding page > # if not fault yet, trigger explicit faultin: > faultin_page(..., FOLL_WRITE, ...) > handle_mm_fault(..., FAULT_FLAG_WRITE) > handle_pte_fault > do_wp_page > (vma->vm_flags & VM_SHARED) wp_page_shared > ... > fault_dirty_shared_page > folio_mark_dirty > a_ops->dirty_folio(), i.e., > filemap_dirty_folio() > # set PG_dirty > folio_test_set_dirty(folio) > # set PAGECACHE_TAG_DIRTY > __folio_mark_dirty > > > # if dio->flags & IOMAP_DIO_DIRTY: > bio_set_pages_dirty > (for each dest page) folio_mark_dirty > a_ops->dirty_folio(), i.e., filemap_dirty_folio() > # set PG_dirty > folio_test_set_dirty(folio) > # set PAGECACHE_TAG_DIRTY > __folio_mark_dirty > ``` > Thanks for this info, Jingbo. > > > > >> ``` > >> > >> The auxiliary write request list was introduced to fix this. > >> > >> I'm not sure if there's an alternative other than the auxiliary list to > >> fix it, e.g. calling folio_wait_writeback() in a_ops->dirty_folio() so > >> that the same folio won't get dirtied when the writeback has not > >> completed yet? > >> > > > > I'm curious how other filesystems solve for this - this seems like a > > generic situation other filesystems would run into as well. > > > > As mentioned above, the writeback path will prevent the duplicate > writeback request on the same page when the first writeback IO has not > completed yet. > > Sorry for the noise... > > -- > Thanks, > Jingbo
On Tue, Nov 12, 2024 at 1:25 AM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: > > Hi Joanne, > > On 11/8/24 7:56 AM, Joanne Koong wrote: > > Currently, we allocate and copy data to a temporary folio when > > handling writeback in order to mitigate the following deadlock scenario > > that may arise if reclaim waits on writeback to complete: > > * single-threaded FUSE server is in the middle of handling a request > > that needs a memory allocation > > * memory allocation triggers direct reclaim > > * direct reclaim waits on a folio under writeback > > * the FUSE server can't write back the folio since it's stuck in > > direct reclaim > > > > To work around this, we allocate a temporary folio and copy over the > > original folio to the temporary folio so that writeback can be > > immediately cleared on the original folio. This additionally requires us > > to maintain an internal rb tree to keep track of writeback state on the > > temporary folios. > > > > A recent change prevents reclaim logic from waiting on writeback for > > folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > > This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > > will prevent FUSE folios from running into the reclaim deadlock described > > above) and removes the temporary folio + extra copying and the internal > > rb tree. > > > > fio benchmarks -- > > (using averages observed from 10 runs, throwing away outliers) > > > > Setup: > > sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > > ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > > > > fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > > --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > > > > bs = 1k 4k 1M > > Before 351 MiB/s 1818 MiB/s 1851 MiB/s > > After 341 MiB/s 2246 MiB/s 2685 MiB/s > > % diff -3% 23% 45% > > > > Signed-off-by: Joanne Koong <joannelkoong@gmail.com> > > I think there are some places checking or waiting for writeback could be > reconsidered if they are still needed or not after we dropping the temp > page design. > > As they are inherited from the original implementation, at least they > are harmless. I think they could be remained in this patch, and could > be cleaned up later if really needed. > Thank you for the thorough inspection! > > > @@ -891,7 +813,7 @@ static int fuse_do_readfolio(struct file *file, struct folio *folio) > > * have writeback that extends beyond the lifetime of the folio. So > > * make sure we read a properly synced folio. > > */ > > - fuse_wait_on_folio_writeback(inode, folio); > > + folio_wait_writeback(folio); > > I doubt if wait-on-writeback is needed here, as now page cache won't be > freed until the writeback IO completes. > > The routine attempts to free page cache, e.g. invalidate_inode_pages2() > (generally called by distributed filesystems when the file content has > been modified from remote) or truncate_inode_pages() (called from > truncate(2) or inode eviction) will wait for writeback completion (if > any) before freeing page. > > Thus I don't think there's any possibility that .read_folio() or > .readahead() will be called when the writeback has not completed. > Great point. I'll remove this line and the comment above it too. > > > @@ -1172,7 +1093,7 @@ static ssize_t fuse_send_write_pages(struct fuse_io_args *ia, > > int err; > > > > for (i = 0; i < ap->num_folios; i++) > > - fuse_wait_on_folio_writeback(inode, ap->folios[i]); > > + folio_wait_writeback(ap->folios[i]); > > Ditto. > Why did we need this fuse_wait_on_folio_writeback() even when we had the temp pages? If I'm understanding it correctly, fuse_send_write_pages() is only called for the writethrough case (by fuse_perform_write()), so these folios would never even be under writeback, no? > > > > static void fuse_writepage_args_page_fill(struct fuse_writepage_args *wpa, struct folio *folio, > > - struct folio *tmp_folio, uint32_t folio_index) > > + uint32_t folio_index) > > { > > struct inode *inode = folio->mapping->host; > > struct fuse_args_pages *ap = &wpa->ia.ap; > > > > - folio_copy(tmp_folio, folio); > > - > > - ap->folios[folio_index] = tmp_folio; > > + folio_get(folio); > > I still think this folio_get() here is harmless but redundant. > > Ditto page cache won't be freed before writeback completes. > > Besides, other .writepages() implementaions e.g. iomap_writepages() also > doen't get the refcount when constructing the writeback IO. > Point taken. I'll remove this then, since other .writepages() also don't obtain a refcount. > > > @@ -2481,7 +2200,7 @@ static int fuse_write_begin(struct file *file, struct address_space *mapping, > > if (IS_ERR(folio)) > > goto error; > > > > - fuse_wait_on_page_writeback(mapping->host, folio->index); > > + folio_wait_writeback(folio); > > I also doubt if wait_on_writeback() is needed here, as now there won't > be duplicate writeback IOs for the same page. What prevents there from being a duplicate writeback write for the same page? This is the path I'm looking at: ksys_write() vfs_write() new_sync_write() op->write_iter() fuse_file_write_iter() fuse_cache_write_iter() generic_file_write_iter() __generic_file_write_iter() generic_perform_write() op->write_begin() fuse_write_begin() but I'm not seeing where there's anything that prevents a duplicate write from happening. > > > > @@ -2545,13 +2264,11 @@ static int fuse_launder_folio(struct folio *folio) > > { > > int err = 0; > > if (folio_clear_dirty_for_io(folio)) { > > - struct inode *inode = folio->mapping->host; > > - > > /* Serialize with pending writeback for the same page */ > > - fuse_wait_on_page_writeback(inode, folio->index); > > + folio_wait_writeback(folio); > > I think folio_wait_writeback() is unneeded after dropping the temp page > copying. This is introduced in commit > 3993382bb3198cc5e263c3519418e716bd57b056 ("fuse: launder page should > wait for page writeback") since .launder_page() could be called when the > previous writeback of the same page has not completed yet. Since now we > won't clear PG_writeback until the writeback completes, .launder_page() > won't be called on the same page when the corresponding writeback IO is > still inflight. > Nice catch, I'll remove this in v4. Thanks, Joanne > > -- > Thanks, > Jingbo
On 11/14/24 8:39 AM, Joanne Koong wrote: > On Tue, Nov 12, 2024 at 1:25 AM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: >> >> Hi Joanne, >> >> On 11/8/24 7:56 AM, Joanne Koong wrote: >>> Currently, we allocate and copy data to a temporary folio when >>> handling writeback in order to mitigate the following deadlock scenario >>> that may arise if reclaim waits on writeback to complete: >>> * single-threaded FUSE server is in the middle of handling a request >>> that needs a memory allocation >>> * memory allocation triggers direct reclaim >>> * direct reclaim waits on a folio under writeback >>> * the FUSE server can't write back the folio since it's stuck in >>> direct reclaim >>> >>> To work around this, we allocate a temporary folio and copy over the >>> original folio to the temporary folio so that writeback can be >>> immediately cleared on the original folio. This additionally requires us >>> to maintain an internal rb tree to keep track of writeback state on the >>> temporary folios. >>> >>> A recent change prevents reclaim logic from waiting on writeback for >>> folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. >>> This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which >>> will prevent FUSE folios from running into the reclaim deadlock described >>> above) and removes the temporary folio + extra copying and the internal >>> rb tree. >>> >>> fio benchmarks -- >>> (using averages observed from 10 runs, throwing away outliers) >>> >>> Setup: >>> sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount >>> ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount >>> >>> fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G >>> --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount >>> >>> bs = 1k 4k 1M >>> Before 351 MiB/s 1818 MiB/s 1851 MiB/s >>> After 341 MiB/s 2246 MiB/s 2685 MiB/s >>> % diff -3% 23% 45% >>> >>> Signed-off-by: Joanne Koong <joannelkoong@gmail.com> >> >> I think there are some places checking or waiting for writeback could be >> reconsidered if they are still needed or not after we dropping the temp >> page design. >> >> As they are inherited from the original implementation, at least they >> are harmless. I think they could be remained in this patch, and could >> be cleaned up later if really needed. >> > > Thank you for the thorough inspection! > >> >>> @@ -891,7 +813,7 @@ static int fuse_do_readfolio(struct file *file, struct folio *folio) >>> * have writeback that extends beyond the lifetime of the folio. So >>> * make sure we read a properly synced folio. >>> */ >>> - fuse_wait_on_folio_writeback(inode, folio); >>> + folio_wait_writeback(folio); >> >> I doubt if wait-on-writeback is needed here, as now page cache won't be >> freed until the writeback IO completes. >> >> The routine attempts to free page cache, e.g. invalidate_inode_pages2() >> (generally called by distributed filesystems when the file content has >> been modified from remote) or truncate_inode_pages() (called from >> truncate(2) or inode eviction) will wait for writeback completion (if >> any) before freeing page. >> >> Thus I don't think there's any possibility that .read_folio() or >> .readahead() will be called when the writeback has not completed. >> > > Great point. I'll remove this line and the comment above it too. > >> >>> @@ -1172,7 +1093,7 @@ static ssize_t fuse_send_write_pages(struct fuse_io_args *ia, >>> int err; >>> >>> for (i = 0; i < ap->num_folios; i++) >>> - fuse_wait_on_folio_writeback(inode, ap->folios[i]); >>> + folio_wait_writeback(ap->folios[i]); >> >> Ditto. Actually this is a typo and I originally meant that waiting for writeback in fuse_send_readpages() is unneeded as page cache won't be freed until the writeback IO completes. > - wait_event(fi->page_waitq, !fuse_range_is_writeback(inode, first, last)); > + filemap_fdatawait_range(inode->i_mapping, first, last); In fact the above waiting for writeback in fuse_send_write_pages() is needed. The reason is as follows: >> > > Why did we need this fuse_wait_on_folio_writeback() even when we had > the temp pages? If I'm understanding it correctly, > fuse_send_write_pages() is only called for the writethrough case (by > fuse_perform_write()), so these folios would never even be under > writeback, no? I think mmap write could make the page dirty and the writeback could be triggered then. > >> >> >>> static void fuse_writepage_args_page_fill(struct fuse_writepage_args *wpa, struct folio *folio, >>> - struct folio *tmp_folio, uint32_t folio_index) >>> + uint32_t folio_index) >>> { >>> struct inode *inode = folio->mapping->host; >>> struct fuse_args_pages *ap = &wpa->ia.ap; >>> >>> - folio_copy(tmp_folio, folio); >>> - >>> - ap->folios[folio_index] = tmp_folio; >>> + folio_get(folio); >> >> I still think this folio_get() here is harmless but redundant. >> >> Ditto page cache won't be freed before writeback completes. >> >> Besides, other .writepages() implementaions e.g. iomap_writepages() also >> doen't get the refcount when constructing the writeback IO. >> > > Point taken. I'll remove this then, since other .writepages() also > don't obtain a refcount. > >> >>> @@ -2481,7 +2200,7 @@ static int fuse_write_begin(struct file *file, struct address_space *mapping, >>> if (IS_ERR(folio)) >>> goto error; >>> >>> - fuse_wait_on_page_writeback(mapping->host, folio->index); >>> + folio_wait_writeback(folio); >> >> I also doubt if wait_on_writeback() is needed here, as now there won't >> be duplicate writeback IOs for the same page. > > What prevents there from being a duplicate writeback write for the > same page? This is the path I'm looking at: > > ksys_write() > vfs_write() > new_sync_write() > op->write_iter() > fuse_file_write_iter() > fuse_cache_write_iter() > generic_file_write_iter() > __generic_file_write_iter() > generic_perform_write() > op->write_begin() > fuse_write_begin() > > but I'm not seeing where there's anything that prevents a duplicate > write from happening. I mean there won't be duplicate *writeback* rather than *write* for the same page. You could write the page cache and make it dirty at the time when the writeback for the same page is still on going, as long as we can ensure that even when the page is dirtied again, there won't be a duplicate writeback IO for the same page when the previous writeback IO has not completed yet.
On Wed, Nov 13, 2024 at 5:47 PM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: > > > > On 11/14/24 8:39 AM, Joanne Koong wrote: > > On Tue, Nov 12, 2024 at 1:25 AM Jingbo Xu <jefflexu@linux.alibaba.com> wrote: > >> > >> Hi Joanne, > >> > >> On 11/8/24 7:56 AM, Joanne Koong wrote: > >>> Currently, we allocate and copy data to a temporary folio when > >>> handling writeback in order to mitigate the following deadlock scenario > >>> that may arise if reclaim waits on writeback to complete: > >>> * single-threaded FUSE server is in the middle of handling a request > >>> that needs a memory allocation > >>> * memory allocation triggers direct reclaim > >>> * direct reclaim waits on a folio under writeback > >>> * the FUSE server can't write back the folio since it's stuck in > >>> direct reclaim > >>> > >>> To work around this, we allocate a temporary folio and copy over the > >>> original folio to the temporary folio so that writeback can be > >>> immediately cleared on the original folio. This additionally requires us > >>> to maintain an internal rb tree to keep track of writeback state on the > >>> temporary folios. > >>> > >>> A recent change prevents reclaim logic from waiting on writeback for > >>> folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. > >>> This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which > >>> will prevent FUSE folios from running into the reclaim deadlock described > >>> above) and removes the temporary folio + extra copying and the internal > >>> rb tree. > >>> > >>> fio benchmarks -- > >>> (using averages observed from 10 runs, throwing away outliers) > >>> > >>> Setup: > >>> sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount > >>> ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount > >>> > >>> fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G > >>> --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount > >>> > >>> bs = 1k 4k 1M > >>> Before 351 MiB/s 1818 MiB/s 1851 MiB/s > >>> After 341 MiB/s 2246 MiB/s 2685 MiB/s > >>> % diff -3% 23% 45% > >>> > >>> Signed-off-by: Joanne Koong <joannelkoong@gmail.com> > >> > >> I think there are some places checking or waiting for writeback could be > >> reconsidered if they are still needed or not after we dropping the temp > >> page design. > >> > >> As they are inherited from the original implementation, at least they > >> are harmless. I think they could be remained in this patch, and could > >> be cleaned up later if really needed. > >> > > > > Thank you for the thorough inspection! > > > >> > >>> @@ -891,7 +813,7 @@ static int fuse_do_readfolio(struct file *file, struct folio *folio) > >>> * have writeback that extends beyond the lifetime of the folio. So > >>> * make sure we read a properly synced folio. > >>> */ > >>> - fuse_wait_on_folio_writeback(inode, folio); > >>> + folio_wait_writeback(folio); > >> > >> I doubt if wait-on-writeback is needed here, as now page cache won't be > >> freed until the writeback IO completes. > >> > >> The routine attempts to free page cache, e.g. invalidate_inode_pages2() > >> (generally called by distributed filesystems when the file content has > >> been modified from remote) or truncate_inode_pages() (called from > >> truncate(2) or inode eviction) will wait for writeback completion (if > >> any) before freeing page. > >> > >> Thus I don't think there's any possibility that .read_folio() or > >> .readahead() will be called when the writeback has not completed. > >> > > > > Great point. I'll remove this line and the comment above it too. > > > >> > >>> @@ -1172,7 +1093,7 @@ static ssize_t fuse_send_write_pages(struct fuse_io_args *ia, > >>> int err; > >>> > >>> for (i = 0; i < ap->num_folios; i++) > >>> - fuse_wait_on_folio_writeback(inode, ap->folios[i]); > >>> + folio_wait_writeback(ap->folios[i]); > >> > >> Ditto. > > Actually this is a typo and I originally meant that waiting for > writeback in fuse_send_readpages() is unneeded as page cache won't be > freed until the writeback IO completes. > > > - wait_event(fi->page_waitq, !fuse_range_is_writeback(inode, first, last)); > > + filemap_fdatawait_range(inode->i_mapping, first, last); > Gotcha and agreed. I'll remove this wait from readahead(). > > In fact the above waiting for writeback in fuse_send_write_pages() is > needed. The reason is as follows: > > >> > > > > Why did we need this fuse_wait_on_folio_writeback() even when we had > > the temp pages? If I'm understanding it correctly, > > fuse_send_write_pages() is only called for the writethrough case (by > > fuse_perform_write()), so these folios would never even be under > > writeback, no? > > I think mmap write could make the page dirty and the writeback could be > triggered then. > Ohhh I see, thanks for the explanation. > > > >> > >> > >>> static void fuse_writepage_args_page_fill(struct fuse_writepage_args *wpa, struct folio *folio, > >>> - struct folio *tmp_folio, uint32_t folio_index) > >>> + uint32_t folio_index) > >>> { > >>> struct inode *inode = folio->mapping->host; > >>> struct fuse_args_pages *ap = &wpa->ia.ap; > >>> > >>> - folio_copy(tmp_folio, folio); > >>> - > >>> - ap->folios[folio_index] = tmp_folio; > >>> + folio_get(folio); > >> > >> I still think this folio_get() here is harmless but redundant. > >> > >> Ditto page cache won't be freed before writeback completes. > >> > >> Besides, other .writepages() implementaions e.g. iomap_writepages() also > >> doen't get the refcount when constructing the writeback IO. > >> > > > > Point taken. I'll remove this then, since other .writepages() also > > don't obtain a refcount. > > > >> > >>> @@ -2481,7 +2200,7 @@ static int fuse_write_begin(struct file *file, struct address_space *mapping, > >>> if (IS_ERR(folio)) > >>> goto error; > >>> > >>> - fuse_wait_on_page_writeback(mapping->host, folio->index); > >>> + folio_wait_writeback(folio); > >> > >> I also doubt if wait_on_writeback() is needed here, as now there won't > >> be duplicate writeback IOs for the same page. > > > > What prevents there from being a duplicate writeback write for the > > same page? This is the path I'm looking at: > > > > ksys_write() > > vfs_write() > > new_sync_write() > > op->write_iter() > > fuse_file_write_iter() > > fuse_cache_write_iter() > > generic_file_write_iter() > > __generic_file_write_iter() > > generic_perform_write() > > op->write_begin() > > fuse_write_begin() > > > > but I'm not seeing where there's anything that prevents a duplicate > > write from happening. > > I mean there won't be duplicate *writeback* rather than *write* for the > same page. You could write the page cache and make it dirty at the time > when the writeback for the same page is still on going, as long as we > can ensure that even when the page is dirtied again, there won't be a > duplicate writeback IO for the same page when the previous writeback IO > has not completed yet. > I think we still need this folio_wait_writeback() since we're calling fuse_do_readfolio() and removing the folio_wait_writeback() from fuse_do_readfolio(). else we could read back stale data if the writeback hasn't gone through yet. I think we could probably move the folio_wait_writeback() here in fuse_write_begin() to be right before the fuse_do_readfolio() call and skip waiting on writeback if we hit the "success" gotos. Thanks, Joanne > > > -- > Thanks, > Jingbo
diff --git a/fs/fuse/file.c b/fs/fuse/file.c index 88d0946b5bc9..f8719d8c56ca 100644 --- a/fs/fuse/file.c +++ b/fs/fuse/file.c @@ -415,89 +415,11 @@ u64 fuse_lock_owner_id(struct fuse_conn *fc, fl_owner_t id) struct fuse_writepage_args { struct fuse_io_args ia; - struct rb_node writepages_entry; struct list_head queue_entry; - struct fuse_writepage_args *next; struct inode *inode; struct fuse_sync_bucket *bucket; }; -static struct fuse_writepage_args *fuse_find_writeback(struct fuse_inode *fi, - pgoff_t idx_from, pgoff_t idx_to) -{ - struct rb_node *n; - - n = fi->writepages.rb_node; - - while (n) { - struct fuse_writepage_args *wpa; - pgoff_t curr_index; - - wpa = rb_entry(n, struct fuse_writepage_args, writepages_entry); - WARN_ON(get_fuse_inode(wpa->inode) != fi); - curr_index = wpa->ia.write.in.offset >> PAGE_SHIFT; - if (idx_from >= curr_index + wpa->ia.ap.num_folios) - n = n->rb_right; - else if (idx_to < curr_index) - n = n->rb_left; - else - return wpa; - } - return NULL; -} - -/* - * Check if any page in a range is under writeback - */ -static bool fuse_range_is_writeback(struct inode *inode, pgoff_t idx_from, - pgoff_t idx_to) -{ - struct fuse_inode *fi = get_fuse_inode(inode); - bool found; - - if (RB_EMPTY_ROOT(&fi->writepages)) - return false; - - spin_lock(&fi->lock); - found = fuse_find_writeback(fi, idx_from, idx_to); - spin_unlock(&fi->lock); - - return found; -} - -static inline bool fuse_page_is_writeback(struct inode *inode, pgoff_t index) -{ - return fuse_range_is_writeback(inode, index, index); -} - -/* - * Wait for page writeback to be completed. - * - * Since fuse doesn't rely on the VM writeback tracking, this has to - * use some other means. - */ -static void fuse_wait_on_page_writeback(struct inode *inode, pgoff_t index) -{ - struct fuse_inode *fi = get_fuse_inode(inode); - - wait_event(fi->page_waitq, !fuse_page_is_writeback(inode, index)); -} - -static inline bool fuse_folio_is_writeback(struct inode *inode, - struct folio *folio) -{ - pgoff_t last = folio_next_index(folio) - 1; - return fuse_range_is_writeback(inode, folio_index(folio), last); -} - -static void fuse_wait_on_folio_writeback(struct inode *inode, - struct folio *folio) -{ - struct fuse_inode *fi = get_fuse_inode(inode); - - wait_event(fi->page_waitq, !fuse_folio_is_writeback(inode, folio)); -} - /* * Wait for all pending writepages on the inode to finish. * @@ -891,7 +813,7 @@ static int fuse_do_readfolio(struct file *file, struct folio *folio) * have writeback that extends beyond the lifetime of the folio. So * make sure we read a properly synced folio. */ - fuse_wait_on_folio_writeback(inode, folio); + folio_wait_writeback(folio); attr_ver = fuse_get_attr_version(fm->fc); @@ -1003,16 +925,15 @@ static void fuse_send_readpages(struct fuse_io_args *ia, struct file *file) static void fuse_readahead(struct readahead_control *rac) { struct inode *inode = rac->mapping->host; - struct fuse_inode *fi = get_fuse_inode(inode); struct fuse_conn *fc = get_fuse_conn(inode); unsigned int max_pages, nr_pages; - pgoff_t first = readahead_index(rac); - pgoff_t last = first + readahead_count(rac) - 1; + loff_t first = readahead_pos(rac); + loff_t last = first + readahead_length(rac) - 1; if (fuse_is_bad(inode)) return; - wait_event(fi->page_waitq, !fuse_range_is_writeback(inode, first, last)); + filemap_fdatawait_range(inode->i_mapping, first, last); max_pages = min_t(unsigned int, fc->max_pages, fc->max_read / PAGE_SIZE); @@ -1172,7 +1093,7 @@ static ssize_t fuse_send_write_pages(struct fuse_io_args *ia, int err; for (i = 0; i < ap->num_folios; i++) - fuse_wait_on_folio_writeback(inode, ap->folios[i]); + folio_wait_writeback(ap->folios[i]); fuse_write_args_fill(ia, ff, pos, count); ia->write.in.flags = fuse_write_flags(iocb); @@ -1622,7 +1543,7 @@ ssize_t fuse_direct_io(struct fuse_io_priv *io, struct iov_iter *iter, return res; } } - if (!cuse && fuse_range_is_writeback(inode, idx_from, idx_to)) { + if (!cuse && filemap_range_has_writeback(mapping, pos, (pos + count - 1))) { if (!write) inode_lock(inode); fuse_sync_writes(inode); @@ -1824,8 +1745,10 @@ static void fuse_writepage_free(struct fuse_writepage_args *wpa) if (wpa->bucket) fuse_sync_bucket_dec(wpa->bucket); - for (i = 0; i < ap->num_folios; i++) + for (i = 0; i < ap->num_folios; i++) { + folio_end_writeback(ap->folios[i]); folio_put(ap->folios[i]); + } fuse_file_put(wpa->ia.ff, false); @@ -1838,7 +1761,7 @@ static void fuse_writepage_finish_stat(struct inode *inode, struct folio *folio) struct backing_dev_info *bdi = inode_to_bdi(inode); dec_wb_stat(&bdi->wb, WB_WRITEBACK); - node_stat_sub_folio(folio, NR_WRITEBACK_TEMP); + node_stat_sub_folio(folio, NR_WRITEBACK); wb_writeout_inc(&bdi->wb); } @@ -1861,7 +1784,6 @@ static void fuse_send_writepage(struct fuse_mount *fm, __releases(fi->lock) __acquires(fi->lock) { - struct fuse_writepage_args *aux, *next; struct fuse_inode *fi = get_fuse_inode(wpa->inode); struct fuse_write_in *inarg = &wpa->ia.write.in; struct fuse_args *args = &wpa->ia.ap.args; @@ -1898,19 +1820,8 @@ __acquires(fi->lock) out_free: fi->writectr--; - rb_erase(&wpa->writepages_entry, &fi->writepages); fuse_writepage_finish(wpa); spin_unlock(&fi->lock); - - /* After rb_erase() aux request list is private */ - for (aux = wpa->next; aux; aux = next) { - next = aux->next; - aux->next = NULL; - fuse_writepage_finish_stat(aux->inode, - aux->ia.ap.folios[0]); - fuse_writepage_free(aux); - } - fuse_writepage_free(wpa); spin_lock(&fi->lock); } @@ -1938,43 +1849,6 @@ __acquires(fi->lock) } } -static struct fuse_writepage_args *fuse_insert_writeback(struct rb_root *root, - struct fuse_writepage_args *wpa) -{ - pgoff_t idx_from = wpa->ia.write.in.offset >> PAGE_SHIFT; - pgoff_t idx_to = idx_from + wpa->ia.ap.num_folios - 1; - struct rb_node **p = &root->rb_node; - struct rb_node *parent = NULL; - - WARN_ON(!wpa->ia.ap.num_folios); - while (*p) { - struct fuse_writepage_args *curr; - pgoff_t curr_index; - - parent = *p; - curr = rb_entry(parent, struct fuse_writepage_args, - writepages_entry); - WARN_ON(curr->inode != wpa->inode); - curr_index = curr->ia.write.in.offset >> PAGE_SHIFT; - - if (idx_from >= curr_index + curr->ia.ap.num_folios) - p = &(*p)->rb_right; - else if (idx_to < curr_index) - p = &(*p)->rb_left; - else - return curr; - } - - rb_link_node(&wpa->writepages_entry, parent, p); - rb_insert_color(&wpa->writepages_entry, root); - return NULL; -} - -static void tree_insert(struct rb_root *root, struct fuse_writepage_args *wpa) -{ - WARN_ON(fuse_insert_writeback(root, wpa)); -} - static void fuse_writepage_end(struct fuse_mount *fm, struct fuse_args *args, int error) { @@ -1994,41 +1868,6 @@ static void fuse_writepage_end(struct fuse_mount *fm, struct fuse_args *args, if (!fc->writeback_cache) fuse_invalidate_attr_mask(inode, FUSE_STATX_MODIFY); spin_lock(&fi->lock); - rb_erase(&wpa->writepages_entry, &fi->writepages); - while (wpa->next) { - struct fuse_mount *fm = get_fuse_mount(inode); - struct fuse_write_in *inarg = &wpa->ia.write.in; - struct fuse_writepage_args *next = wpa->next; - - wpa->next = next->next; - next->next = NULL; - tree_insert(&fi->writepages, next); - - /* - * Skip fuse_flush_writepages() to make it easy to crop requests - * based on primary request size. - * - * 1st case (trivial): there are no concurrent activities using - * fuse_set/release_nowrite. Then we're on safe side because - * fuse_flush_writepages() would call fuse_send_writepage() - * anyway. - * - * 2nd case: someone called fuse_set_nowrite and it is waiting - * now for completion of all in-flight requests. This happens - * rarely and no more than once per page, so this should be - * okay. - * - * 3rd case: someone (e.g. fuse_do_setattr()) is in the middle - * of fuse_set_nowrite..fuse_release_nowrite section. The fact - * that fuse_set_nowrite returned implies that all in-flight - * requests were completed along with all of their secondary - * requests. Further primary requests are blocked by negative - * writectr. Hence there cannot be any in-flight requests and - * no invocations of fuse_writepage_end() while we're in - * fuse_set_nowrite..fuse_release_nowrite section. - */ - fuse_send_writepage(fm, next, inarg->offset + inarg->size); - } fi->writectr--; fuse_writepage_finish(wpa); spin_unlock(&fi->lock); @@ -2115,19 +1954,18 @@ static void fuse_writepage_add_to_bucket(struct fuse_conn *fc, } static void fuse_writepage_args_page_fill(struct fuse_writepage_args *wpa, struct folio *folio, - struct folio *tmp_folio, uint32_t folio_index) + uint32_t folio_index) { struct inode *inode = folio->mapping->host; struct fuse_args_pages *ap = &wpa->ia.ap; - folio_copy(tmp_folio, folio); - - ap->folios[folio_index] = tmp_folio; + folio_get(folio); + ap->folios[folio_index] = folio; ap->descs[folio_index].offset = 0; ap->descs[folio_index].length = PAGE_SIZE; inc_wb_stat(&inode_to_bdi(inode)->wb, WB_WRITEBACK); - node_stat_add_folio(tmp_folio, NR_WRITEBACK_TEMP); + node_stat_add_folio(folio, NR_WRITEBACK); } static struct fuse_writepage_args *fuse_writepage_args_setup(struct folio *folio, @@ -2162,18 +2000,12 @@ static int fuse_writepage_locked(struct folio *folio) struct fuse_inode *fi = get_fuse_inode(inode); struct fuse_writepage_args *wpa; struct fuse_args_pages *ap; - struct folio *tmp_folio; struct fuse_file *ff; - int error = -ENOMEM; - - tmp_folio = folio_alloc(GFP_NOFS | __GFP_HIGHMEM, 0); - if (!tmp_folio) - goto err; + int error = -EIO; - error = -EIO; ff = fuse_write_file_get(fi); if (!ff) - goto err_nofile; + goto err; wpa = fuse_writepage_args_setup(folio, ff); error = -ENOMEM; @@ -2184,22 +2016,17 @@ static int fuse_writepage_locked(struct folio *folio) ap->num_folios = 1; folio_start_writeback(folio); - fuse_writepage_args_page_fill(wpa, folio, tmp_folio, 0); + fuse_writepage_args_page_fill(wpa, folio, 0); spin_lock(&fi->lock); - tree_insert(&fi->writepages, wpa); list_add_tail(&wpa->queue_entry, &fi->queued_writes); fuse_flush_writepages(inode); spin_unlock(&fi->lock); - folio_end_writeback(folio); - return 0; err_writepage_args: fuse_file_put(ff, false); -err_nofile: - folio_put(tmp_folio); err: mapping_set_error(folio->mapping, error); return error; @@ -2209,7 +2036,6 @@ struct fuse_fill_wb_data { struct fuse_writepage_args *wpa; struct fuse_file *ff; struct inode *inode; - struct folio **orig_folios; unsigned int max_folios; }; @@ -2244,69 +2070,11 @@ static void fuse_writepages_send(struct fuse_fill_wb_data *data) struct fuse_writepage_args *wpa = data->wpa; struct inode *inode = data->inode; struct fuse_inode *fi = get_fuse_inode(inode); - int num_folios = wpa->ia.ap.num_folios; - int i; spin_lock(&fi->lock); list_add_tail(&wpa->queue_entry, &fi->queued_writes); fuse_flush_writepages(inode); spin_unlock(&fi->lock); - - for (i = 0; i < num_folios; i++) - folio_end_writeback(data->orig_folios[i]); -} - -/* - * Check under fi->lock if the page is under writeback, and insert it onto the - * rb_tree if not. Otherwise iterate auxiliary write requests, to see if there's - * one already added for a page at this offset. If there's none, then insert - * this new request onto the auxiliary list, otherwise reuse the existing one by - * swapping the new temp page with the old one. - */ -static bool fuse_writepage_add(struct fuse_writepage_args *new_wpa, - struct folio *folio) -{ - struct fuse_inode *fi = get_fuse_inode(new_wpa->inode); - struct fuse_writepage_args *tmp; - struct fuse_writepage_args *old_wpa; - struct fuse_args_pages *new_ap = &new_wpa->ia.ap; - - WARN_ON(new_ap->num_folios != 0); - new_ap->num_folios = 1; - - spin_lock(&fi->lock); - old_wpa = fuse_insert_writeback(&fi->writepages, new_wpa); - if (!old_wpa) { - spin_unlock(&fi->lock); - return true; - } - - for (tmp = old_wpa->next; tmp; tmp = tmp->next) { - pgoff_t curr_index; - - WARN_ON(tmp->inode != new_wpa->inode); - curr_index = tmp->ia.write.in.offset >> PAGE_SHIFT; - if (curr_index == folio->index) { - WARN_ON(tmp->ia.ap.num_folios != 1); - swap(tmp->ia.ap.folios[0], new_ap->folios[0]); - break; - } - } - - if (!tmp) { - new_wpa->next = old_wpa->next; - old_wpa->next = new_wpa; - } - - spin_unlock(&fi->lock); - - if (tmp) { - fuse_writepage_finish_stat(new_wpa->inode, - folio); - fuse_writepage_free(new_wpa); - } - - return false; } static bool fuse_writepage_need_send(struct fuse_conn *fc, struct folio *folio, @@ -2315,15 +2083,6 @@ static bool fuse_writepage_need_send(struct fuse_conn *fc, struct folio *folio, { WARN_ON(!ap->num_folios); - /* - * Being under writeback is unlikely but possible. For example direct - * read to an mmaped fuse file will set the page dirty twice; once when - * the pages are faulted with get_user_pages(), and then after the read - * completed. - */ - if (fuse_folio_is_writeback(data->inode, folio)) - return true; - /* Reached max pages */ if (ap->num_folios == fc->max_pages) return true; @@ -2333,7 +2092,7 @@ static bool fuse_writepage_need_send(struct fuse_conn *fc, struct folio *folio, return true; /* Discontinuity */ - if (data->orig_folios[ap->num_folios - 1]->index + 1 != folio_index(folio)) + if (ap->folios[ap->num_folios - 1]->index + 1 != folio_index(folio)) return true; /* Need to grow the pages array? If so, did the expansion fail? */ @@ -2352,7 +2111,6 @@ static int fuse_writepages_fill(struct folio *folio, struct inode *inode = data->inode; struct fuse_inode *fi = get_fuse_inode(inode); struct fuse_conn *fc = get_fuse_conn(inode); - struct folio *tmp_folio; int err; if (!data->ff) { @@ -2367,54 +2125,23 @@ static int fuse_writepages_fill(struct folio *folio, data->wpa = NULL; } - err = -ENOMEM; - tmp_folio = folio_alloc(GFP_NOFS | __GFP_HIGHMEM, 0); - if (!tmp_folio) - goto out_unlock; - - /* - * The page must not be redirtied until the writeout is completed - * (i.e. userspace has sent a reply to the write request). Otherwise - * there could be more than one temporary page instance for each real - * page. - * - * This is ensured by holding the page lock in page_mkwrite() while - * checking fuse_page_is_writeback(). We already hold the page lock - * since clear_page_dirty_for_io() and keep it held until we add the - * request to the fi->writepages list and increment ap->num_folios. - * After this fuse_page_is_writeback() will indicate that the page is - * under writeback, so we can release the page lock. - */ if (data->wpa == NULL) { err = -ENOMEM; wpa = fuse_writepage_args_setup(folio, data->ff); - if (!wpa) { - folio_put(tmp_folio); + if (!wpa) goto out_unlock; - } fuse_file_get(wpa->ia.ff); data->max_folios = 1; ap = &wpa->ia.ap; } folio_start_writeback(folio); - fuse_writepage_args_page_fill(wpa, folio, tmp_folio, ap->num_folios); - data->orig_folios[ap->num_folios] = folio; + fuse_writepage_args_page_fill(wpa, folio, ap->num_folios); err = 0; - if (data->wpa) { - /* - * Protected by fi->lock against concurrent access by - * fuse_page_is_writeback(). - */ - spin_lock(&fi->lock); - ap->num_folios++; - spin_unlock(&fi->lock); - } else if (fuse_writepage_add(wpa, folio)) { + ap->num_folios++; + if (!data->wpa) data->wpa = wpa; - } else { - folio_end_writeback(folio); - } out_unlock: folio_unlock(folio); @@ -2441,13 +2168,6 @@ static int fuse_writepages(struct address_space *mapping, data.wpa = NULL; data.ff = NULL; - err = -ENOMEM; - data.orig_folios = kcalloc(fc->max_pages, - sizeof(struct folio *), - GFP_NOFS); - if (!data.orig_folios) - goto out; - err = write_cache_pages(mapping, wbc, fuse_writepages_fill, &data); if (data.wpa) { WARN_ON(!data.wpa->ia.ap.num_folios); @@ -2456,7 +2176,6 @@ static int fuse_writepages(struct address_space *mapping, if (data.ff) fuse_file_put(data.ff, false); - kfree(data.orig_folios); out: return err; } @@ -2481,7 +2200,7 @@ static int fuse_write_begin(struct file *file, struct address_space *mapping, if (IS_ERR(folio)) goto error; - fuse_wait_on_page_writeback(mapping->host, folio->index); + folio_wait_writeback(folio); if (folio_test_uptodate(folio) || len >= folio_size(folio)) goto success; @@ -2545,13 +2264,11 @@ static int fuse_launder_folio(struct folio *folio) { int err = 0; if (folio_clear_dirty_for_io(folio)) { - struct inode *inode = folio->mapping->host; - /* Serialize with pending writeback for the same page */ - fuse_wait_on_page_writeback(inode, folio->index); + folio_wait_writeback(folio); err = fuse_writepage_locked(folio); if (!err) - fuse_wait_on_page_writeback(inode, folio->index); + folio_wait_writeback(folio); } return err; } @@ -2595,7 +2312,7 @@ static vm_fault_t fuse_page_mkwrite(struct vm_fault *vmf) return VM_FAULT_NOPAGE; } - fuse_wait_on_folio_writeback(inode, folio); + folio_wait_writeback(folio); return VM_FAULT_LOCKED; } @@ -3413,9 +3130,12 @@ static const struct address_space_operations fuse_file_aops = { void fuse_init_file_inode(struct inode *inode, unsigned int flags) { struct fuse_inode *fi = get_fuse_inode(inode); + struct fuse_conn *fc = get_fuse_conn(inode); inode->i_fop = &fuse_file_operations; inode->i_data.a_ops = &fuse_file_aops; + if (fc->writeback_cache) + mapping_set_writeback_may_block(&inode->i_data); INIT_LIST_HEAD(&fi->write_files); INIT_LIST_HEAD(&fi->queued_writes); @@ -3423,7 +3143,6 @@ void fuse_init_file_inode(struct inode *inode, unsigned int flags) fi->iocachectr = 0; init_waitqueue_head(&fi->page_waitq); init_waitqueue_head(&fi->direct_io_waitq); - fi->writepages = RB_ROOT; if (IS_ENABLED(CONFIG_FUSE_DAX)) fuse_dax_inode_init(inode, flags);
Currently, we allocate and copy data to a temporary folio when handling writeback in order to mitigate the following deadlock scenario that may arise if reclaim waits on writeback to complete: * single-threaded FUSE server is in the middle of handling a request that needs a memory allocation * memory allocation triggers direct reclaim * direct reclaim waits on a folio under writeback * the FUSE server can't write back the folio since it's stuck in direct reclaim To work around this, we allocate a temporary folio and copy over the original folio to the temporary folio so that writeback can be immediately cleared on the original folio. This additionally requires us to maintain an internal rb tree to keep track of writeback state on the temporary folios. A recent change prevents reclaim logic from waiting on writeback for folios whose mappings have the AS_WRITEBACK_MAY_BLOCK flag set in it. This commit sets AS_WRITEBACK_MAY_BLOCK on FUSE inode mappings (which will prevent FUSE folios from running into the reclaim deadlock described above) and removes the temporary folio + extra copying and the internal rb tree. fio benchmarks -- (using averages observed from 10 runs, throwing away outliers) Setup: sudo mount -t tmpfs -o size=30G tmpfs ~/tmp_mount ./libfuse/build/example/passthrough_ll -o writeback -o max_threads=4 -o source=~/tmp_mount ~/fuse_mount fio --name=writeback --ioengine=sync --rw=write --bs={1k,4k,1M} --size=2G --numjobs=2 --ramp_time=30 --group_reporting=1 --directory=/root/fuse_mount bs = 1k 4k 1M Before 351 MiB/s 1818 MiB/s 1851 MiB/s After 341 MiB/s 2246 MiB/s 2685 MiB/s % diff -3% 23% 45% Signed-off-by: Joanne Koong <joannelkoong@gmail.com> --- fs/fuse/file.c | 339 +++++-------------------------------------------- 1 file changed, 29 insertions(+), 310 deletions(-)