Changes in write.c

8e5419d6 - nfs: Add missing release on error in nfs_lock_and_join_requests()

Wed, 02 Apr 2025 11:02:40 +0000

nfs: Add missing release on error in nfs_lock_and_join_requests()Call nfs_release_request() on this error path before returning.Fixes: c3f2235782c3 ("nfs: fold nfs_folio_find_and_lock_request into nfs_lock_and_join_requests")Signed-off-by: Dan Carpenter Link: https://lore.kernel.org/r/3aaaa3d5-1c8a-41e4-98c7-717801ddd171@stanley.mountainSigned-off-by: Trond Myklebust List of files: /linux-6.15/fs/nfs/write.c

86e00412 - nfs: cache all open LOCALIO nfsd_file(s) in client

Sat, 16 Nov 2024 01:40:57 +0000

nfs: cache all open LOCALIO nfsd_file(s) in clientThis commit switches from leaning heavily on NFSD's filecache (interms of GC'd nfsd_files) back to caching nfsd_files in theclient. A later commit will add the callback mechanism needed toallow NFSD to force the NFS client to cleanup all cached nfsd_files.Add nfs_fh_localio_init() and 'struct nfs_fh_localio' to cache openednfsd_file(s) (both a RO and RW nfsd_file is able to be opened andcached for a given nfs_fh).Update nfs_local_open_fh() to cache the nfsd_file once it is openedusing __nfs_local_open_fh().Introduce nfs_close_local_fh() to clear the cached open nfsd_files andcall nfs_to_nfsd_file_put_local().Refcounting is such that:- nfs_local_open_fh() is paired with nfs_close_local_fh().- __nfs_local_open_fh() is paired with nfs_to_nfsd_file_put_local().- nfs_local_file_get() is paired with nfs_local_file_put().Signed-off-by: Mike Snitzer Reviewed-by: Jeff Layton Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

66f9dac9 - Revert "nfs: don't reuse partially completed requests in nfs_lock_and_join_requests"

Tue, 05 Nov 2024 02:09:21 +0000

Revert "nfs: don't reuse partially completed requests in nfs_lock_and_join_requests"This reverts commit b571cfcb9dcac187c6d967987792d37cb0688610.This patch appears to assume that if one request is complete, then theothers will complete too before unlocking. That is not a validassumption, since other requests could hit a non-fatal error or a shortwrite that would cause them not to complete.Reported-by: Igor Raits Link: https://bugzilla.kernel.org/show_bug.cgi?id=219508Fixes: b571cfcb9dca ("nfs: don't reuse partially completed requests in nfs_lock_and_join_requests")Signed-off-by: Trond Myklebust List of files: /linux-6.15/fs/nfs/write.c

8f52caf9 - Revert "fs: nfs: fix missing refcnt by replacing folio_set_private by folio_attach_private"

Tue, 05 Nov 2024 02:01:43 +0000

Revert "fs: nfs: fix missing refcnt by replacing folio_set_private by folio_attach_private"This reverts commit 03e02b94171b1985dd0aa184296fe94425b855a3.As was pointed out during code review, there is no need to usefolio_attach_private()/folio_detach_private() when a refcount to thefolio is already carried by the struct nfs_page.Signed-off-by: Trond Myklebust List of files: /linux-6.15/fs/nfs/write.c

fa88a7d6 - nfs: enable localio for non-pNFS IO

Thu, 05 Sep 2024 19:09:54 +0000

nfs: enable localio for non-pNFS IOTry a local open of the file being written to, and if it succeeds,then use localio to issue IO.Signed-off-by: Trond Myklebust Signed-off-by: Mike Snitzer Reviewed-by: NeilBrown Reviewed-by: Jeff Layton Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

70ba381e - nfs: add LOCALIO support

Thu, 05 Sep 2024 19:09:53 +0000

nfs: add LOCALIO supportAdd client support for bypassing NFS for localhost reads, writes, andcommits. This is only useful when the client and the server arerunning on the same host.nfs_local_probe() is stubbed out, later commits will enable client andserver handshake via a Linux-only LOCALIO auxiliary RPC protocol.This has dynamic binding with the nfsd module (via nfs_localio modulewhich is part of nfs_common). LOCALIO will only work if nfsd isalready loaded.The "localio_enabled" nfs kernel module parameter can be used todisable and enable the ability to use LOCALIO support.CONFIG_NFS_LOCALIO enables NFS client support for LOCALIO.Lastly, LOCALIO uses an nfsd_file to initiate all IO. To make properuse of nfsd_file (and nfsd's filecache) its lifetime (duration beforenfsd_file_put is called) must extend until after commit, read andwrite operations. So rather than immediately drop the nfsd_filereference in nfs_local_open_fh(), that doesn't happen untilnfs_local_pgio_release() for read/write and not untilnfs_local_release_commit_data() for commit. The same applies to thereference held on nfsd's nn->nfsd_serv. Both objects' lifetimes andassociated references are managed through calls tonfs_to->nfsd_file_put_local().Signed-off-by: Weston Andros Adamson Signed-off-by: Trond Myklebust Co-developed-by: Mike Snitzer Signed-off-by: Mike Snitzer Signed-off-by: NeilBrown # nfs_open_local_fhReviewed-by: Jeff Layton Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

df24c483 - nfs: pass struct nfsd_file to nfs_init_pgio and nfs_init_commit

Thu, 05 Sep 2024 19:09:52 +0000

nfs: pass struct nfsd_file to nfs_init_pgio and nfs_init_commitThe nfsd_file will be passed, in future commits, by callersthat enable LOCALIO support (for both regular NFS and pNFS IO).[Derived from patch authored by Weston Andros Adamson, but switched from passing struct file to struct nfsd_file]Signed-off-by: Mike Snitzer Reviewed-by: NeilBrown Reviewed-by: Jeff Layton Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

dfb07e99 - nfs: add 'noalignwrite' option for lock-less 'lost writes' prevention

Wed, 24 Jul 2024 11:07:12 +0000

nfs: add 'noalignwrite' option for lock-less 'lost writes' preventionThere are some applications that write to predefined non-overlappingfile offsets from multiple clients and therefore don't need to rely onfile locking. However, if these applications want non-aligned offsetsand sizes they need to either use locks or risk data corruption, as theNFS client defaults to extending writes to whole pages.This commit adds a new mount option `noalignwrite`, which allows to turnthat off and avoid the need of locking, as long as these applicationsdon't overlap on offsets.Signed-off-by: Dan Aloni Reviewed-by: Jeff Layton Reviewed-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

03e02b94 - fs: nfs: fix missing refcnt by replacing folio_set_private by folio_attach_private

Fri, 30 Aug 2024 03:27:47 +0000

fs: nfs: fix missing refcnt by replacing folio_set_private by folio_attach_privateThis patch is inspired by a code review of fs codes which aims atfolio's extra refcnt that could introduce unwanted behavious whenjudging refcnt, such as[1].That is, the folio passed tomapping_evict_folio carries the refcnts from find_lock_entries,page_cache, corresponding to PTEs and folio's private if has. However,current code doesn't take the refcnt for folio's private which couldhave mapping_evict_folio miss the one to only PTE and lead tocall filemap_release_folio wrongly.[1]long mapping_evict_folio(struct address_space *mapping, struct folio *folio){...//current code will misjudge here if there is one pte on the folio whichis be deemed as the one as folio's private if (folio_ref_count(folio) > folio_nr_pages(folio) + folio_has_private(folio) + 1) return 0; if (!filemap_release_folio(folio, 0)) return 0; return remove_mapping(mapping, folio);}Signed-off-by: Zhaoyang Huang Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

fada32ed - nfs: pass explicit offset/count to trace events

Thu, 11 Jul 2024 07:17:02 +0000

nfs: pass explicit offset/count to trace eventsnfs_folio_length is unsafe to use without having the folio locked and acheck for a NULL ->f_mapping that protects against truncations and canlead to kernel crashes. E.g. when running xfstests generic/065 withall nfs trace points enabled.Follow the model of the XFS trace points and pass in an explіcit offsetand length. This has the additional benefit that these values canbe more accurate as some of the users touch partial folio ranges.Fixes: eb5654b3b89d ("NFS: Enable tracing of nfs_invalidate_folio() and nfs_launder_folio()")Reported-by: Chuck Lever Signed-off-by: Christoph Hellwig Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

39c910a4 - nfs: do not extend writes to the entire folio

Fri, 05 Jul 2024 05:42:51 +0000

nfs: do not extend writes to the entire folionfs_update_folio has code to extend a write to the entire page undercertain conditions. With the support for large folios this nowsuddenly extents to the variable sized and potentially much larger folio.Add code to limit the extension to the page boundaries of the start andend of the write, which matches the historic expecation and the codecomments.Fixes: b73fe2dd6cd5 ("nfs: add support for large folios")Signed-off-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Reviewed-by: Benjamin Coddington Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

b571cfcb - nfs: don't reuse partially completed requests in nfs_lock_and_join_requests

Mon, 01 Jul 2024 05:26:54 +0000

nfs: don't reuse partially completed requests in nfs_lock_and_join_requestsWhen NFS requests are split into sub-requests, nfs_inode_remove_requestcalls nfs_page_group_sync_on_bit to set PG_REMOVE on this sub-request andonly completes the head requests once PG_REMOVE is set on all requests.This means that when nfs_lock_and_join_requests sees a PG_REMOVE bit, I/Oon the request is in progress and has partially completed. If such arequest is returned to nfs_try_to_update_request, it could be extendedwith the newly dirtied region and I/O for the combined range will bere-scheduled, leading to extra I/O.Change the logic to instead restart the search for a request when anyPG_REMOVE bit is set, as the completion handler will remove the requestas soon as it can take the page group lock. This not only avoidextending the I/O but also does the right thing for the callers thatwant to cancel or flush the request.Signed-off-by: Christoph Hellwig Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

f1b7c755 - nfs: move nfs_wait_on_request to write.c

Mon, 01 Jul 2024 05:26:53 +0000

nfs: move nfs_wait_on_request to write.cnfs_wait_on_request is now only used in write.c. Move it thereand mark it static.Signed-off-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

25edbcac - nfs: fold nfs_page_group_lock_subrequests into nfs_lock_and_join_requests

Mon, 01 Jul 2024 05:26:52 +0000

nfs: fold nfs_page_group_lock_subrequests into nfs_lock_and_join_requestsFold nfs_page_group_lock_subrequests into nfs_lock_and_join_requests toprepare for future changes to this code, and move the helpers to write.cas well.Signed-off-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

c3f22357 - nfs: fold nfs_folio_find_and_lock_request into nfs_lock_and_join_requests

Mon, 01 Jul 2024 05:26:51 +0000

nfs: fold nfs_folio_find_and_lock_request into nfs_lock_and_join_requestsFold nfs_folio_find_and_lock_request into the only caller to preparefor changes to this code.Signed-off-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

9eb7c484 - nfs: simplify nfs_folio_find_and_lock_request

Mon, 01 Jul 2024 05:26:50 +0000

nfs: simplify nfs_folio_find_and_lock_requestnfs_folio_find_and_lock_request and the nfs_page_group_lock_head helpercalled by it spend quite some effort to deal with head vs subrequests.But given that only the head request can be stashed in the folio privatedata, non of that is required.Fold the locking logic from nfs_page_group_lock_head intonfs_folio_find_and_lock_request and simplify the result based on theinvariant that we always find the head request in the folio private data.Signed-off-by: Christoph Hellwig Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

02e61ec1 - nfs: remove nfs_folio_private_request

Mon, 01 Jul 2024 05:26:49 +0000

nfs: remove nfs_folio_private_requestnfs_folio_private_request is a trivial wrapper around, which itself hasfallen out of favor and has been replaced with plain ->privatedereferences in recent folio conversions. Do the same for nfs.Signed-off-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

7e8e78a0 - nfs: remove dead code for the old swap over NFS implementation

Mon, 01 Jul 2024 05:26:48 +0000

nfs: remove dead code for the old swap over NFS implementationRemove the code testing folio_test_swapcache either explicitly orimplicitly in pagemap.h headers, as is now handled using the direct I/Opath and not the buffered I/O path that these helpers are located in.Signed-off-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

2f1f3104 - nfs: Block on write congestion

Mon, 01 Jul 2024 10:50:48 +0000

nfs: Block on write congestionCommit 6df25e58532b ("nfs: remove reliance on bdi congestion")introduced NFS-private solution for limiting number of writesoutstanding against a particular server. Unlike previous bdi congestionthis algorithm actually works and limits number of outstanding writebackpages to nfs_congestion_kb which scales with amount of client's memoryand is capped at 256 MB. As a result some workloads such as randombuffered writes over NFS got slower (from ~170 MB/s to ~126 MB/s). Thefio command to reproduce is:fio --direct=0 --ioengine=sync --thread --invalidate=1 --group_reporting=1 --runtime=300 --fallocate=posix --ramp_time=10 --new_group --rw=randwrite --size=64256m --numjobs=4 --bs=4k --fsync_on_close=1 --end_fsync=1This happens because the client sends ~256 MB worth of dirty pages tothe server and any further background writeback request is ignored untilthe number of writeback pages gets below the threshold of 192 MB. By thetime this happens and clients decides to trigger another round ofwriteback, the server often has no pages to write and the disk is idle.To fix this problem and make the client react faster to eased congestionof the server by blocking waiting for congestion to resolve instead ofaborting writeback. This improves the random 4k buffered writethroughput to 184 MB/s.Reviewed-by: Sagi Grimberg Reviewed-by: Jeff Layton Signed-off-by: Jan Kara Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c

37d4159d - nfs: Drop pointless check from nfs_commit_release_pages()

Mon, 01 Jul 2024 10:50:46 +0000

nfs: Drop pointless check from nfs_commit_release_pages()nfss->writeback is updated only when we are ending page writeback and atthat moment we also clear nfss->write_congested. So there's no point inrechecking congestion state in nfs_commit_release_pages(). Drop thepointless check.Reviewed-by: Sagi Grimberg Reviewed-by: Jeff Layton Signed-off-by: Jan Kara Signed-off-by: Anna Schumaker List of files: /linux-6.15/fs/nfs/write.c