Cachefiles keeps track of how much space is available on the backing filesystem and refuses new writes permission to start if there isn't enough (we especially don't want ENOSPC happening). It also tracks the amount of data pending in DIO writes (cache->b_writing) and reduces the amount of free space available by this amount before deciding if it can set up a new write. However, the old fscache I/O API was very much page-granularity dependent and, as such, cachefiles's cache->bshift was meant to be a multiplier to get from PAGE_SIZE to block size (ie. a blocksize of 512 would give a shift of 3 for a 4KiB page) - and this was incorrectly being used to turn the number of bytes in a DIO write into a number of blocks, leading to a massive over estimation of the amount of data in flight. Fix this by changing cache->bshift to be a multiplier from bytes to blocksize and deal with quantities of blocks, not quantities of pages. Fix also the rounding in the calculation in cachefiles_write() which needs a "- 1" inserting. Fixes: 047487c947e8 ("cachefiles: Implement the I/O routines") Signed-off-by: David Howells <dhowells@xxxxxxxxxx> cc: linux-cachefs@xxxxxxxxxx --- fs/cachefiles/cache.c | 7 ++----- fs/cachefiles/internal.h | 2 +- fs/cachefiles/io.c | 2 +- 3 files changed, 4 insertions(+), 7 deletions(-) diff --git a/fs/cachefiles/cache.c b/fs/cachefiles/cache.c index ce4d4785003c..1e9c71666c6a 100644 --- a/fs/cachefiles/cache.c +++ b/fs/cachefiles/cache.c @@ -84,9 +84,7 @@ int cachefiles_add_cache(struct cachefiles_cache *cache) goto error_unsupported; cache->bsize = stats.f_bsize; - cache->bshift = 0; - if (stats.f_bsize < PAGE_SIZE) - cache->bshift = PAGE_SHIFT - ilog2(stats.f_bsize); + cache->bshift = ilog2(stats.f_bsize); _debug("blksize %u (shift %u)", cache->bsize, cache->bshift); @@ -106,7 +104,6 @@ int cachefiles_add_cache(struct cachefiles_cache *cache) (unsigned long long) cache->fcull, (unsigned long long) cache->fstop); - stats.f_blocks >>= cache->bshift; do_div(stats.f_blocks, 100); cache->bstop = stats.f_blocks * cache->bstop_percent; cache->bcull = stats.f_blocks * cache->bcull_percent; @@ -209,7 +206,7 @@ int cachefiles_has_space(struct cachefiles_cache *cache, return ret; } - b_avail = stats.f_bavail >> cache->bshift; + b_avail = stats.f_bavail; b_writing = atomic_long_read(&cache->b_writing); if (b_avail > b_writing) b_avail -= b_writing; diff --git a/fs/cachefiles/internal.h b/fs/cachefiles/internal.h index 8dd54d9375b6..c793d33b0224 100644 --- a/fs/cachefiles/internal.h +++ b/fs/cachefiles/internal.h @@ -86,7 +86,7 @@ struct cachefiles_cache { unsigned bcull_percent; /* when to start culling (% blocks) */ unsigned bstop_percent; /* when to stop allocating (% blocks) */ unsigned bsize; /* cache's block size */ - unsigned bshift; /* min(ilog2(PAGE_SIZE / bsize), 0) */ + unsigned bshift; /* ilog2(bsize) */ uint64_t frun; /* when to stop culling */ uint64_t fcull; /* when to start culling */ uint64_t fstop; /* when to stop allocating */ diff --git a/fs/cachefiles/io.c b/fs/cachefiles/io.c index 60b1eac2ce78..04eb52736990 100644 --- a/fs/cachefiles/io.c +++ b/fs/cachefiles/io.c @@ -264,7 +264,7 @@ static int cachefiles_write(struct netfs_cache_resources *cres, ki->term_func = term_func; ki->term_func_priv = term_func_priv; ki->was_async = true; - ki->b_writing = (len + (1 << cache->bshift)) >> cache->bshift; + ki->b_writing = (len + (1 << cache->bshift) - 1) >> cache->bshift; if (ki->term_func) ki->iocb.ki_complete = cachefiles_write_complete;