From 0e4d464cda4c5996402343d4c9e2b6ceec716f93 Mon Sep 17 00:00:00 2001 From: David Howells Date: Fri, 5 Jan 2024 14:57:14 +0000 Subject: [PATCH 1/6] netfs: Mark netfs_unbuffered_write_iter_locked() static Mark netfs_unbuffered_write_iter_locked() static as it's only called from the file in which it is defined. Signed-off-by: David Howells cc: Jeff Layton cc: linux-cachefs@redhat.com cc: linux-fsdevel@vger.kernel.org cc: linux-mm@kvack.org --- fs/netfs/direct_write.c | 4 ++-- fs/netfs/internal.h | 6 ------ 2 files changed, 2 insertions(+), 8 deletions(-) diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c index aad05f2349a4..b9cbfd6a8a01 100644 --- a/fs/netfs/direct_write.c +++ b/fs/netfs/direct_write.c @@ -27,8 +27,8 @@ static void netfs_cleanup_dio_write(struct netfs_io_request *wreq) * Perform an unbuffered write where we may have to do an RMW operation on an * encrypted file. This can also be used for direct I/O writes. */ -ssize_t netfs_unbuffered_write_iter_locked(struct kiocb *iocb, struct iov_iter *iter, - struct netfs_group *netfs_group) +static ssize_t netfs_unbuffered_write_iter_locked(struct kiocb *iocb, struct iov_iter *iter, + struct netfs_group *netfs_group) { struct netfs_io_request *wreq; unsigned long long start = iocb->ki_pos; diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h index d2d63120ac60..a6dfc8888377 100644 --- a/fs/netfs/internal.h +++ b/fs/netfs/internal.h @@ -26,12 +26,6 @@ void netfs_rreq_unlock_folios(struct netfs_io_request *rreq); int netfs_prefetch_for_write(struct file *file, struct folio *folio, size_t offset, size_t len); -/* - * direct_write.c - */ -ssize_t netfs_unbuffered_write_iter_locked(struct kiocb *iocb, struct iov_iter *iter, - struct netfs_group *netfs_group); - /* * io.c */ From 4088e389476e3baababf9b22f34b9d8b3e557344 Mon Sep 17 00:00:00 2001 From: David Howells Date: Fri, 5 Jan 2024 14:55:52 +0000 Subject: [PATCH 2/6] netfs: Count DIO writes Provide a counter for DIO writes to match that for DIO reads. Signed-off-by: David Howells cc: Jeff Layton cc: linux-cachefs@redhat.com cc: linux-fsdevel@vger.kernel.org cc: linux-mm@kvack.org --- fs/netfs/direct_write.c | 1 + fs/netfs/internal.h | 1 + fs/netfs/stats.c | 11 +++++++---- 3 files changed, 9 insertions(+), 4 deletions(-) diff --git a/fs/netfs/direct_write.c b/fs/netfs/direct_write.c index b9cbfd6a8a01..60a40d293c87 100644 --- a/fs/netfs/direct_write.c +++ b/fs/netfs/direct_write.c @@ -140,6 +140,7 @@ ssize_t netfs_unbuffered_write_iter(struct kiocb *iocb, struct iov_iter *from) _enter("%llx,%zx,%llx", iocb->ki_pos, iov_iter_count(from), i_size_read(inode)); trace_netfs_write_iter(iocb, from); + netfs_stat(&netfs_n_rh_dio_write); ret = netfs_start_io_direct(inode); if (ret < 0) diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h index a6dfc8888377..3f9620d0fa63 100644 --- a/fs/netfs/internal.h +++ b/fs/netfs/internal.h @@ -104,6 +104,7 @@ int netfs_end_writethrough(struct netfs_io_request *wreq, struct kiocb *iocb); */ #ifdef CONFIG_NETFS_STATS extern atomic_t netfs_n_rh_dio_read; +extern atomic_t netfs_n_rh_dio_write; extern atomic_t netfs_n_rh_readahead; extern atomic_t netfs_n_rh_readpage; extern atomic_t netfs_n_rh_rreq; diff --git a/fs/netfs/stats.c b/fs/netfs/stats.c index 15fd5c3f0f39..42db36528d92 100644 --- a/fs/netfs/stats.c +++ b/fs/netfs/stats.c @@ -10,6 +10,7 @@ #include "internal.h" atomic_t netfs_n_rh_dio_read; +atomic_t netfs_n_rh_dio_write; atomic_t netfs_n_rh_readahead; atomic_t netfs_n_rh_readpage; atomic_t netfs_n_rh_rreq; @@ -37,14 +38,13 @@ atomic_t netfs_n_wh_write_failed; int netfs_stats_show(struct seq_file *m, void *v) { - seq_printf(m, "Netfs : DR=%u RA=%u RP=%u WB=%u WBZ=%u rr=%u sr=%u\n", + seq_printf(m, "Netfs : DR=%u DW=%u RA=%u RP=%u WB=%u WBZ=%u\n", atomic_read(&netfs_n_rh_dio_read), + atomic_read(&netfs_n_rh_dio_write), atomic_read(&netfs_n_rh_readahead), atomic_read(&netfs_n_rh_readpage), atomic_read(&netfs_n_rh_write_begin), - atomic_read(&netfs_n_rh_write_zskip), - atomic_read(&netfs_n_rh_rreq), - atomic_read(&netfs_n_rh_sreq)); + atomic_read(&netfs_n_rh_write_zskip)); seq_printf(m, "Netfs : ZR=%u sh=%u sk=%u\n", atomic_read(&netfs_n_rh_zero), atomic_read(&netfs_n_rh_short_read), @@ -66,6 +66,9 @@ int netfs_stats_show(struct seq_file *m, void *v) atomic_read(&netfs_n_wh_write), atomic_read(&netfs_n_wh_write_done), atomic_read(&netfs_n_wh_write_failed)); + seq_printf(m, "Netfs : rr=%u sr=%u\n", + atomic_read(&netfs_n_rh_rreq), + atomic_read(&netfs_n_rh_sreq)); return fscache_stats_show(m); } EXPORT_SYMBOL(netfs_stats_show); From 92a714d727ec9e7ccfcc7432d348aba730145914 Mon Sep 17 00:00:00 2001 From: David Howells Date: Thu, 4 Jan 2024 15:52:11 +0000 Subject: [PATCH 3/6] netfs: Fix interaction between write-streaming and cachefiles culling An issue can occur between write-streaming (storing dirty data in partial non-uptodate pages) and a cachefiles object being culled to make space. The problem occurs because the cache object is only marked in use while there are files open using it. Once it has been released, it can be culled and the cookie marked disabled. At this point, a streaming write is permitted to occur (if the cache is active, we require pages to be prefetched and cached), but the cache can become active again before this gets flushed out - and then two effects can occur: (1) The cache may be asked to write out a region that's less than its DIO block size (assumed by cachefiles to be PAGE_SIZE) - and this causes one of two debugging statements to be emitted. (2) netfs_how_to_modify() gets confused because it sees a page that isn't allowed to be non-uptodate being uptodate and tries to prefetch it - leading to a warning that PG_fscache is set twice. Fix this by the following means: (1) Add a netfs_inode flag to disallow write-streaming to an inode and set it if we ever do local caching of that inode. It remains set for the lifetime of that inode - even if the cookie becomes disabled. (2) If the no-write-streaming flag is set, then make netfs_how_to_modify() always want to prefetch instead. (3) If netfs_how_to_modify() decides it wants to prefetch a folio, but that folio has write-streamed data in it, then it requires the folio be flushed first. (4) Export a counter of the number of times we wanted to prefetch a non-uptodate page, but found it had write-streamed data in it. (5) Export a counter of the number of times we cancelled a write to the cache because it didn't DIO align and remove the debug statements. Reported-by: Marc Dionne Signed-off-by: David Howells cc: Jeff Layton cc: linux-cachefs@redhat.com cc: linux-erofs@lists.ozlabs.org cc: linux-fsdevel@vger.kernel.org cc: linux-mm@kvack.org --- fs/cachefiles/io.c | 12 ++++++------ fs/netfs/buffered_write.c | 22 +++++++++++++++++++--- fs/netfs/fscache_stats.c | 9 ++++++--- fs/netfs/internal.h | 1 + fs/netfs/stats.c | 6 ++++-- include/linux/fscache-cache.h | 3 +++ include/linux/netfs.h | 1 + 7 files changed, 40 insertions(+), 14 deletions(-) diff --git a/fs/cachefiles/io.c b/fs/cachefiles/io.c index 7529b40bc95a..3eec26967437 100644 --- a/fs/cachefiles/io.c +++ b/fs/cachefiles/io.c @@ -528,12 +528,12 @@ int __cachefiles_prepare_write(struct cachefiles_object *object, /* Round to DIO size */ start = round_down(*_start, PAGE_SIZE); - if (start != *_start) { - kleave(" = -ENOBUFS [down]"); - return -ENOBUFS; - } - if (*_len > upper_len) { - kleave(" = -ENOBUFS [up]"); + if (start != *_start || *_len > upper_len) { + /* Probably asked to cache a streaming write written into the + * pagecache when the cookie was temporarily out of service to + * culling. + */ + fscache_count_dio_misfit(); return -ENOBUFS; } diff --git a/fs/netfs/buffered_write.c b/fs/netfs/buffered_write.c index 08f28800232c..6cd8f7422e9a 100644 --- a/fs/netfs/buffered_write.c +++ b/fs/netfs/buffered_write.c @@ -80,10 +80,19 @@ static enum netfs_how_to_modify netfs_how_to_modify(struct netfs_inode *ctx, return NETFS_WHOLE_FOLIO_MODIFY; if (file->f_mode & FMODE_READ) - return NETFS_JUST_PREFETCH; + goto no_write_streaming; + if (test_bit(NETFS_ICTX_NO_WRITE_STREAMING, &ctx->flags)) + goto no_write_streaming; - if (netfs_is_cache_enabled(ctx)) - return NETFS_JUST_PREFETCH; + if (netfs_is_cache_enabled(ctx)) { + /* We don't want to get a streaming write on a file that loses + * caching service temporarily because the backing store got + * culled. + */ + if (!test_bit(NETFS_ICTX_NO_WRITE_STREAMING, &ctx->flags)) + set_bit(NETFS_ICTX_NO_WRITE_STREAMING, &ctx->flags); + goto no_write_streaming; + } if (!finfo) return NETFS_STREAMING_WRITE; @@ -95,6 +104,13 @@ static enum netfs_how_to_modify netfs_how_to_modify(struct netfs_inode *ctx, if (offset == finfo->dirty_offset + finfo->dirty_len) return NETFS_STREAMING_WRITE_CONT; return NETFS_FLUSH_CONTENT; + +no_write_streaming: + if (finfo) { + netfs_stat(&netfs_n_wh_wstream_conflict); + return NETFS_FLUSH_CONTENT; + } + return NETFS_JUST_PREFETCH; } /* diff --git a/fs/netfs/fscache_stats.c b/fs/netfs/fscache_stats.c index aad812ead398..add21abdf713 100644 --- a/fs/netfs/fscache_stats.c +++ b/fs/netfs/fscache_stats.c @@ -48,13 +48,15 @@ atomic_t fscache_n_no_create_space; EXPORT_SYMBOL(fscache_n_no_create_space); atomic_t fscache_n_culled; EXPORT_SYMBOL(fscache_n_culled); +atomic_t fscache_n_dio_misfit; +EXPORT_SYMBOL(fscache_n_dio_misfit); /* * display the general statistics */ int fscache_stats_show(struct seq_file *m) { - seq_puts(m, "FS-Cache statistics\n"); + seq_puts(m, "-- FS-Cache statistics --\n"); seq_printf(m, "Cookies: n=%d v=%d vcol=%u voom=%u\n", atomic_read(&fscache_n_cookies), atomic_read(&fscache_n_volumes), @@ -93,8 +95,9 @@ int fscache_stats_show(struct seq_file *m) atomic_read(&fscache_n_no_create_space), atomic_read(&fscache_n_culled)); - seq_printf(m, "IO : rd=%u wr=%u\n", + seq_printf(m, "IO : rd=%u wr=%u mis=%u\n", atomic_read(&fscache_n_read), - atomic_read(&fscache_n_write)); + atomic_read(&fscache_n_write), + atomic_read(&fscache_n_dio_misfit)); return 0; } diff --git a/fs/netfs/internal.h b/fs/netfs/internal.h index 3f9620d0fa63..ec7045d24400 100644 --- a/fs/netfs/internal.h +++ b/fs/netfs/internal.h @@ -123,6 +123,7 @@ extern atomic_t netfs_n_rh_write_begin; extern atomic_t netfs_n_rh_write_done; extern atomic_t netfs_n_rh_write_failed; extern atomic_t netfs_n_rh_write_zskip; +extern atomic_t netfs_n_wh_wstream_conflict; extern atomic_t netfs_n_wh_upload; extern atomic_t netfs_n_wh_upload_done; extern atomic_t netfs_n_wh_upload_failed; diff --git a/fs/netfs/stats.c b/fs/netfs/stats.c index 42db36528d92..deeba9f9dcf5 100644 --- a/fs/netfs/stats.c +++ b/fs/netfs/stats.c @@ -29,6 +29,7 @@ atomic_t netfs_n_rh_write_begin; atomic_t netfs_n_rh_write_done; atomic_t netfs_n_rh_write_failed; atomic_t netfs_n_rh_write_zskip; +atomic_t netfs_n_wh_wstream_conflict; atomic_t netfs_n_wh_upload; atomic_t netfs_n_wh_upload_done; atomic_t netfs_n_wh_upload_failed; @@ -66,9 +67,10 @@ int netfs_stats_show(struct seq_file *m, void *v) atomic_read(&netfs_n_wh_write), atomic_read(&netfs_n_wh_write_done), atomic_read(&netfs_n_wh_write_failed)); - seq_printf(m, "Netfs : rr=%u sr=%u\n", + seq_printf(m, "Netfs : rr=%u sr=%u wsc=%u\n", atomic_read(&netfs_n_rh_rreq), - atomic_read(&netfs_n_rh_sreq)); + atomic_read(&netfs_n_rh_sreq), + atomic_read(&netfs_n_wh_wstream_conflict)); return fscache_stats_show(m); } EXPORT_SYMBOL(netfs_stats_show); diff --git a/include/linux/fscache-cache.h b/include/linux/fscache-cache.h index a174cedf4d90..bdf7f3eddf0a 100644 --- a/include/linux/fscache-cache.h +++ b/include/linux/fscache-cache.h @@ -189,17 +189,20 @@ extern atomic_t fscache_n_write; extern atomic_t fscache_n_no_write_space; extern atomic_t fscache_n_no_create_space; extern atomic_t fscache_n_culled; +extern atomic_t fscache_n_dio_misfit; #define fscache_count_read() atomic_inc(&fscache_n_read) #define fscache_count_write() atomic_inc(&fscache_n_write) #define fscache_count_no_write_space() atomic_inc(&fscache_n_no_write_space) #define fscache_count_no_create_space() atomic_inc(&fscache_n_no_create_space) #define fscache_count_culled() atomic_inc(&fscache_n_culled) +#define fscache_count_dio_misfit() atomic_inc(&fscache_n_dio_misfit) #else #define fscache_count_read() do {} while(0) #define fscache_count_write() do {} while(0) #define fscache_count_no_write_space() do {} while(0) #define fscache_count_no_create_space() do {} while(0) #define fscache_count_culled() do {} while(0) +#define fscache_count_dio_misfit() do {} while(0) #endif #endif /* _LINUX_FSCACHE_CACHE_H */ diff --git a/include/linux/netfs.h b/include/linux/netfs.h index d3bac60fcd6f..100cbb261269 100644 --- a/include/linux/netfs.h +++ b/include/linux/netfs.h @@ -142,6 +142,7 @@ struct netfs_inode { #define NETFS_ICTX_ODIRECT 0 /* The file has DIO in progress */ #define NETFS_ICTX_UNBUFFERED 1 /* I/O should not use the pagecache */ #define NETFS_ICTX_WRITETHROUGH 2 /* Write-through caching */ +#define NETFS_ICTX_NO_WRITE_STREAMING 3 /* Don't engage in write-streaming */ }; /* From 807c6d09cc99cbdf9933edfadcbaa8f0b856848d Mon Sep 17 00:00:00 2001 From: David Howells Date: Fri, 5 Jan 2024 22:03:58 +0000 Subject: [PATCH 4/6] netfs: Fix the loop that unmarks folios after writing to the cache In the loop in netfs_rreq_unmark_after_write() that removes the PG_fscache from folios after they've been written to the cache, as soon as we remove the mark from a multipage folio, it can get split - and then we might see a fragment of folio again. Guard against this by advancing the 'unlocked' tracker to the index of the last page in the folio to avoid a double removal of the PG_fscache mark. Reported-by: Marc Dionne Signed-off-by: David Howells cc: Matthew Wilcox cc: linux-afs@lists.infradead.org cc: linux-cachefs@redhat.com cc: linux-fsdevel@vger.kernel.org cc: linux-mm@kvack.org --- fs/netfs/buffered_write.c | 1 + fs/netfs/io.c | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) diff --git a/fs/netfs/buffered_write.c b/fs/netfs/buffered_write.c index 6cd8f7422e9a..0b2b7a60dabc 100644 --- a/fs/netfs/buffered_write.c +++ b/fs/netfs/buffered_write.c @@ -698,6 +698,7 @@ static void netfs_pages_written_back(struct netfs_io_request *wreq) end_wb: if (folio_test_fscache(folio)) folio_end_fscache(folio); + xas_advance(&xas, folio_next_index(folio) - 1); folio_end_writeback(folio); } diff --git a/fs/netfs/io.c b/fs/netfs/io.c index 5b5af96cd4b9..4309edf33862 100644 --- a/fs/netfs/io.c +++ b/fs/netfs/io.c @@ -126,7 +126,7 @@ static void netfs_rreq_unmark_after_write(struct netfs_io_request *rreq, */ if (have_unlocked && folio_index(folio) <= unlocked) continue; - unlocked = folio_index(folio); + unlocked = folio_next_index(folio) - 1; trace_netfs_folio(folio, netfs_folio_trace_end_copy); folio_end_fscache(folio); have_unlocked = true; From 3d1d4aa0cc13b1883a5a56c945837a2e0ecb5143 Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Jan 2024 10:02:55 +0000 Subject: [PATCH 5/6] cachefiles: Fix signed/unsigned mixup In __cachefiles_prepare_write(), the start and pos variables were made unsigned 64-bit so that the casts in the checking could be got rid of - which should be fine since absolute file offsets can't be negative, except that an error code may be obtained from vfs_llseek(), which *would* be negative. This breaks the error check. Fix this for now by reverting pos and start to be signed and putting back the casts. Unfortunately, the error value checks cannot be replaced with IS_ERR_VALUE() as long might be 32-bits. Fixes: 7097c96411d2 ("cachefiles: Fix __cachefiles_prepare_write()") Reported-by: Simon Horman Reported-by: kernel test robot Closes: https://lore.kernel.org/oe-kbuild-all/202401071152.DbKqMQMu-lkp@intel.com/ Signed-off-by: David Howells Reviewed-by: Simon Horman Reviewed-by: Gao Xiang cc: Yiqun Leng cc: Jia Zhu cc: Jeff Layton cc: linux-cachefs@redhat.com cc: linux-erofs@lists.ozlabs.org cc: linux-fsdevel@vger.kernel.org cc: linux-mm@kvack.org --- fs/cachefiles/io.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/fs/cachefiles/io.c b/fs/cachefiles/io.c index 3eec26967437..9a2cb2868e90 100644 --- a/fs/cachefiles/io.c +++ b/fs/cachefiles/io.c @@ -522,7 +522,7 @@ int __cachefiles_prepare_write(struct cachefiles_object *object, bool no_space_allocated_yet) { struct cachefiles_cache *cache = object->volume->cache; - unsigned long long start = *_start, pos; + loff_t start = *_start, pos; size_t len = *_len; int ret; @@ -556,7 +556,7 @@ int __cachefiles_prepare_write(struct cachefiles_object *object, cachefiles_trace_seek_error); return pos; } - if (pos >= start + *_len) + if ((u64)pos >= (u64)start + *_len) goto check_space; /* Unallocated region */ /* We have a block that's at least partially filled - if we're low on @@ -575,7 +575,7 @@ int __cachefiles_prepare_write(struct cachefiles_object *object, cachefiles_trace_seek_error); return pos; } - if (pos >= start + *_len) + if ((u64)pos >= (u64)start + *_len) return 0; /* Fully allocated */ /* Partially allocated, but insufficient space: cull. */ From e2bdb5272f4314256f51d91eee7babcae58b194b Mon Sep 17 00:00:00 2001 From: David Howells Date: Mon, 8 Jan 2024 21:30:49 +0000 Subject: [PATCH 6/6] netfs: Fix wrong #ifdef hiding wait netfs_writepages_begin() has the wait on the fscache folio conditional on CONFIG_NETFS_FSCACHE - which doesn't exist. Fix it to be conditional on CONFIG_FSCACHE instead. Fixes: 62c3b7481b9a ("netfs: Provide a writepages implementation") Reported-by: Lukas Bulwahn Signed-off-by: David Howells cc: Matthew Wilcox cc: linux-afs@lists.infradead.org cc: linux-cachefs@redhat.com cc: linux-fsdevel@vger.kernel.org cc: linux-mm@kvack.org Link: https://lore.kernel.org/r/20240109083257.GK132648@kernel.org/ --- fs/netfs/buffered_write.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/netfs/buffered_write.c b/fs/netfs/buffered_write.c index 0b2b7a60dabc..de517ca70d91 100644 --- a/fs/netfs/buffered_write.c +++ b/fs/netfs/buffered_write.c @@ -1076,7 +1076,7 @@ lock_again: folio_unlock(folio); if (wbc->sync_mode != WB_SYNC_NONE) { folio_wait_writeback(folio); -#ifdef CONFIG_NETFS_FSCACHE +#ifdef CONFIG_FSCACHE folio_wait_fscache(folio); #endif goto lock_again;