| Message ID | f803021382040dba38ce8414ed1db8e400c0cc49.1611160121.git.bcodding@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | NFS client readdir per-page validation | expand |
On Wed, 2021-01-20 at 11:59 -0500, Benjamin Coddington wrote: > Whenever we successfully locate our dir_cookie within the pagecache, > or > finish emitting entries to userspace, update the pagecache cursor. > These > updates provide marker points to validate pagecache pages in a future > patch. How isn't this going to end up subject to the exact same problem that Dave Wysochanski's patchset had? > > Signed-off-by: Benjamin Coddington <bcodding@redhat.com> > --- > fs/nfs/dir.c | 29 +++++++++++++++++++++++++---- > 1 file changed, 25 insertions(+), 4 deletions(-) > > diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c > index 6626aff9f54d..7f6c84c8a412 100644 > --- a/fs/nfs/dir.c > +++ b/fs/nfs/dir.c > @@ -150,6 +150,10 @@ struct nfs_cache_array { > struct nfs_cache_array_entry array[]; > }; > > +static const int cache_entries_per_page = > + (PAGE_SIZE - sizeof(struct nfs_cache_array)) / > + sizeof(struct nfs_cache_array_entry); > + > struct nfs_readdir_descriptor { > struct file *file; > struct page *page; > @@ -251,6 +255,21 @@ static bool nfs_readdir_array_is_full(struct > nfs_cache_array *array) > return array->page_full; > } > > +static void nfs_readdir_set_cursor(struct nfs_readdir_descriptor > *desc, int index) > +{ > + desc->pgc.entry_index = index; > + desc->pgc.index_cookie = desc->dir_cookie; > +} > + > +static void nfs_readdir_cursor_next(struct nfs_dir_page_cursor *pgc, > u64 cookie) > +{ > + pgc->index_cookie = cookie; > + if (++pgc->entry_index == cache_entries_per_page) { > + pgc->entry_index = 0; > + pgc->page_index++; > + } > +} > + > /* > * the caller is responsible for freeing qstr.name > * when called by nfs_readdir_add_to_array, the strings will be > freed in > @@ -424,7 +443,7 @@ static int nfs_readdir_search_for_pos(struct > nfs_cache_array *array, > > index = (unsigned int)diff; > desc->dir_cookie = array->array[index].cookie; > - desc->pgc.entry_index = index; > + nfs_readdir_set_cursor(desc, index); > return 0; > out_eof: > desc->eof = true; > @@ -492,7 +511,7 @@ static int nfs_readdir_search_for_cookie(struct > nfs_cache_array *array, > else > desc->ctx->pos = new_pos; > desc->prev_index = new_pos; > - desc->pgc.entry_index = i; > + nfs_readdir_set_cursor(desc, i); > return 0; > } > } > @@ -519,9 +538,9 @@ static int nfs_readdir_search_array(struct > nfs_readdir_descriptor *desc) > status = nfs_readdir_search_for_cookie(array, desc); > > if (status == -EAGAIN) { > - desc->pgc.index_cookie = array->last_cookie; > + desc->pgc.entry_index = array->size - 1; > + nfs_readdir_cursor_next(&desc->pgc, array- > >last_cookie); > desc->current_index += array->size; > - desc->pgc.page_index++; > } > kunmap_atomic(array); > return status; > @@ -1035,6 +1054,8 @@ static void nfs_do_filldir(struct > nfs_readdir_descriptor *desc) > desc->eof = true; > > kunmap(desc->page); > + desc->pgc.entry_index = i-1; > + nfs_readdir_cursor_next(&desc->pgc, desc->dir_cookie); > dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling ended @ > cookie %llu\n", > (unsigned long long)desc->dir_cookie); > }
On Thu, 2021-01-21 at 20:00 +0000, Trond Myklebust wrote: > On Wed, 2021-01-20 at 11:59 -0500, Benjamin Coddington wrote: > > Whenever we successfully locate our dir_cookie within the > > pagecache, > > or > > finish emitting entries to userspace, update the pagecache cursor. > > These > > updates provide marker points to validate pagecache pages in a > > future > > patch. > > How isn't this going to end up subject to the exact same problem that > Dave Wysochanski's patchset had? IOW: how is this not also making invalid assumptions around page cache layout stability across READDIR calls?
On 21 Jan 2021, at 15:11, Trond Myklebust wrote: > On Thu, 2021-01-21 at 20:00 +0000, Trond Myklebust wrote: >> On Wed, 2021-01-20 at 11:59 -0500, Benjamin Coddington wrote: >>> Whenever we successfully locate our dir_cookie within the >>> pagecache, >>> or >>> finish emitting entries to userspace, update the pagecache cursor. >>> These >>> updates provide marker points to validate pagecache pages in a >>> future >>> patch. >> >> How isn't this going to end up subject to the exact same problem that >> Dave Wysochanski's patchset had? > > IOW: how is this not also making invalid assumptions around page cache > layout stability across READDIR calls? IIRC, Dave's approach was to store the index along with dir_cookie in order to skip having to re-fill the cache to resume a listing. That approach assumed that the index still referred to page data aligned with the current reader's dir_cookie. But in between calls to nfs_readdir() it would be possible for another reader to fill the cache with a different alignment due to directory changes. In which case the index is not going to point to the next entry, we'll either skip entries or repeat them. With this approach, we don't assume this is the case. For every page with data, the alignment of the entries is verified to be in the same position as that reader's last pass through. If not, the page data is discarded and refreshed with a new READDIR op - just like an uncached_readdir path, but we still fill the pagecache. Every READDIR also updates the change_attr, so even if there are cached pages beyond our current location that match our alignment, we detect the case where those pages are actually invalid due to changes on the server, and we re-fill them. Another way to think about this is that instead of trying to cache the complete representation of the directory aligned to the first entry in the pagecache, we're instead just caching the results of any READDIR at convenient offsets in the pagecache for other readers that might follow. The READDIR results are only usable if they match the current version of the directory and their alignment is correct. If you were to lseek to nearly the end of a directory, the first call to nfs_readdir() will end up with results of a READDIR hitting the cache and filling page->index 0. That's ok because any other reader coming along with a different alignment will discard that data and refresh it. We are going to create a performance penalty for two readers that want to regularly fill entries at different alignments, but I think that case is probably fairly rare. I am making a guess, but I think the most common usage of readdir is by readers that want to traverse the entire directory in order. So for that case all the readers benefit from the work of other processes, the cache can be filled in parallel, and readers at the beginning don't prevent readers at the end from filling in entries. We no longer have to worry whether we have enough memory to list a directory, or play games with timeouts. Ben
diff --git a/fs/nfs/dir.c b/fs/nfs/dir.c index 6626aff9f54d..7f6c84c8a412 100644 --- a/fs/nfs/dir.c +++ b/fs/nfs/dir.c @@ -150,6 +150,10 @@ struct nfs_cache_array { struct nfs_cache_array_entry array[]; }; +static const int cache_entries_per_page = + (PAGE_SIZE - sizeof(struct nfs_cache_array)) / + sizeof(struct nfs_cache_array_entry); + struct nfs_readdir_descriptor { struct file *file; struct page *page; @@ -251,6 +255,21 @@ static bool nfs_readdir_array_is_full(struct nfs_cache_array *array) return array->page_full; } +static void nfs_readdir_set_cursor(struct nfs_readdir_descriptor *desc, int index) +{ + desc->pgc.entry_index = index; + desc->pgc.index_cookie = desc->dir_cookie; +} + +static void nfs_readdir_cursor_next(struct nfs_dir_page_cursor *pgc, u64 cookie) +{ + pgc->index_cookie = cookie; + if (++pgc->entry_index == cache_entries_per_page) { + pgc->entry_index = 0; + pgc->page_index++; + } +} + /* * the caller is responsible for freeing qstr.name * when called by nfs_readdir_add_to_array, the strings will be freed in @@ -424,7 +443,7 @@ static int nfs_readdir_search_for_pos(struct nfs_cache_array *array, index = (unsigned int)diff; desc->dir_cookie = array->array[index].cookie; - desc->pgc.entry_index = index; + nfs_readdir_set_cursor(desc, index); return 0; out_eof: desc->eof = true; @@ -492,7 +511,7 @@ static int nfs_readdir_search_for_cookie(struct nfs_cache_array *array, else desc->ctx->pos = new_pos; desc->prev_index = new_pos; - desc->pgc.entry_index = i; + nfs_readdir_set_cursor(desc, i); return 0; } } @@ -519,9 +538,9 @@ static int nfs_readdir_search_array(struct nfs_readdir_descriptor *desc) status = nfs_readdir_search_for_cookie(array, desc); if (status == -EAGAIN) { - desc->pgc.index_cookie = array->last_cookie; + desc->pgc.entry_index = array->size - 1; + nfs_readdir_cursor_next(&desc->pgc, array->last_cookie); desc->current_index += array->size; - desc->pgc.page_index++; } kunmap_atomic(array); return status; @@ -1035,6 +1054,8 @@ static void nfs_do_filldir(struct nfs_readdir_descriptor *desc) desc->eof = true; kunmap(desc->page); + desc->pgc.entry_index = i-1; + nfs_readdir_cursor_next(&desc->pgc, desc->dir_cookie); dfprintk(DIRCACHE, "NFS: nfs_do_filldir() filling ended @ cookie %llu\n", (unsigned long long)desc->dir_cookie); }
Whenever we successfully locate our dir_cookie within the pagecache, or finish emitting entries to userspace, update the pagecache cursor. These updates provide marker points to validate pagecache pages in a future patch. Signed-off-by: Benjamin Coddington <bcodding@redhat.com> --- fs/nfs/dir.c | 29 +++++++++++++++++++++++++---- 1 file changed, 25 insertions(+), 4 deletions(-)