* [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O @ 2025-07-15 3:10 Qi Han via Linux-f2fs-devel 2025-07-15 6:58 ` Chao Yu via Linux-f2fs-devel 2025-07-15 14:28 ` Jens Axboe 0 siblings, 2 replies; 10+ messages in thread From: Qi Han via Linux-f2fs-devel @ 2025-07-15 3:10 UTC (permalink / raw) To: jaegeuk, chao; +Cc: axboe, Qi Han, linux-kernel, linux-f2fs-devel Jens has already completed the development of uncached buffered I/O in git [1], and in f2fs, the feature can be enabled simply by setting the FOP_DONTCACHE flag in f2fs_file_operations. [1] https://lore.kernel.org/all/20241220154831.1086649-10-axboe@kernel.dk/T/#m58520a94b46f543d82db3711453dfc7bb594b2b0 Signed-off-by: Qi Han <hanqi@vivo.com> --- fs/f2fs/file.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 696131e655ed..d8da1fc2febf 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -5425,5 +5425,5 @@ const struct file_operations f2fs_file_operations = { .splice_read = f2fs_file_splice_read, .splice_write = iter_file_splice_write, .fadvise = f2fs_file_fadvise, - .fop_flags = FOP_BUFFER_RASYNC, + .fop_flags = FOP_BUFFER_RASYNC | FOP_DONTCACHE, }; -- 2.48.1 _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-15 3:10 [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O Qi Han via Linux-f2fs-devel @ 2025-07-15 6:58 ` Chao Yu via Linux-f2fs-devel 2025-07-15 8:14 ` hanqi via Linux-f2fs-devel 2025-07-15 14:28 ` Jens Axboe 1 sibling, 1 reply; 10+ messages in thread From: Chao Yu via Linux-f2fs-devel @ 2025-07-15 6:58 UTC (permalink / raw) To: Qi Han, jaegeuk; +Cc: axboe, linux-kernel, linux-f2fs-devel On 7/15/25 11:10, Qi Han wrote: > Jens has already completed the development of uncached buffered I/O > in git [1], and in f2fs, the feature can be enabled simply by setting > the FOP_DONTCACHE flag in f2fs_file_operations. Hi Qi, do you have any numbers of f2fs before/after this change? though I'm not against supporting this feature in f2fs. Thanks, > > [1] > https://lore.kernel.org/all/20241220154831.1086649-10-axboe@kernel.dk/T/#m58520a94b46f543d82db3711453dfc7bb594b2b0 > > Signed-off-by: Qi Han <hanqi@vivo.com> > --- > fs/f2fs/file.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c > index 696131e655ed..d8da1fc2febf 100644 > --- a/fs/f2fs/file.c > +++ b/fs/f2fs/file.c > @@ -5425,5 +5425,5 @@ const struct file_operations f2fs_file_operations = { > .splice_read = f2fs_file_splice_read, > .splice_write = iter_file_splice_write, > .fadvise = f2fs_file_fadvise, > - .fop_flags = FOP_BUFFER_RASYNC, > + .fop_flags = FOP_BUFFER_RASYNC | FOP_DONTCACHE, > }; _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-15 6:58 ` Chao Yu via Linux-f2fs-devel @ 2025-07-15 8:14 ` hanqi via Linux-f2fs-devel 0 siblings, 0 replies; 10+ messages in thread From: hanqi via Linux-f2fs-devel @ 2025-07-15 8:14 UTC (permalink / raw) To: Chao Yu, jaegeuk; +Cc: axboe, linux-kernel, linux-f2fs-devel 在 2025/7/15 14:58, Chao Yu 写道: > On 7/15/25 11:10, Qi Han wrote: >> Jens has already completed the development of uncached buffered I/O >> in git [1], and in f2fs, the feature can be enabled simply by setting >> the FOP_DONTCACHE flag in f2fs_file_operations. > Hi Qi, do you have any numbers of f2fs before/after this change? though > I'm not against supporting this feature in f2fs. > > Thanks, Hi, Chao I have been testing a use case locally, which aligns with Jens' test case [1]. In the read scenario, using uncached buffer I/O results in more stable read performance and a lower load on the background memory reclaim thread (kswapd). However, in the write scenario, it appears that uncached buffer I/O may not be suitable for F2FS. This is because F2FS calls folio_end_writeback in the softirq context, as discussed in [2]. Read test data without using uncached buffer I/O: reading bs 32768, uncached 0 1s: 1856MB/sec, MB=1856 2s: 1907MB/sec, MB=3763 3s: 1830MB/sec, MB=5594 4s: 1745MB/sec, MB=7333 5s: 1829MB/sec, MB=9162 6s: 1903MB/sec, MB=11075 7s: 1878MB/sec, MB=12942 8s: 1763MB/sec, MB=14718 9s: 1845MB/sec, MB=16549 10s: 1915MB/sec, MB=18481 11s: 1831MB/sec, MB=20295 12s: 1750MB/sec, MB=22066 13s: 1787MB/sec, MB=23832 14s: 1913MB/sec, MB=25769 15s: 1898MB/sec, MB=27668 16s: 1795MB/sec, MB=29436 17s: 1812MB/sec, MB=31248 18s: 1890MB/sec, MB=33139 19s: 1880MB/sec, MB=35020 20s: 1754MB/sec, MB=36810 08:36:26 UID PID %usr %system %guest %wait %CPU CPU Command 08:36:27 0 93 0.00 0.00 0.00 0.00 0.00 7 kswapd0 08:36:28 0 93 0.00 0.00 0.00 0.00 0.00 7 kswapd0 08:36:29 0 93 0.00 0.00 0.00 0.00 0.00 7 kswapd0 08:36:30 0 93 0.00 56.00 0.00 0.00 56.00 7 kswapd0 08:36:31 0 93 0.00 73.00 0.00 0.00 73.00 7 kswapd0 08:36:32 0 93 0.00 83.00 0.00 0.00 83.00 7 kswapd0 08:36:33 0 93 0.00 75.00 0.00 0.00 75.00 7 kswapd0 08:36:34 0 93 0.00 81.00 0.00 0.00 81.00 7 kswapd0 08:36:35 0 93 0.00 54.00 0.00 1.00 54.00 2 kswapd0 08:36:36 0 93 0.00 61.00 0.00 0.00 61.00 0 kswapd0 08:36:37 0 93 0.00 68.00 0.00 0.00 68.00 7 kswapd0 08:36:38 0 93 0.00 53.00 0.00 0.00 53.00 2 kswapd0 08:36:39 0 93 0.00 82.00 0.00 0.00 82.00 7 kswapd0 08:36:40 0 93 0.00 77.00 0.00 0.00 77.00 1 kswapd0 08:36:41 0 93 0.00 74.00 0.00 1.00 74.00 7 kswapd0 08:36:42 0 93 0.00 71.00 0.00 0.00 71.00 7 kswapd0 08:36:43 0 93 0.00 78.00 0.00 0.00 78.00 7 kswapd0 08:36:44 0 93 0.00 85.00 0.00 0.00 85.00 7 kswapd0 08:36:45 0 93 0.00 83.00 0.00 0.00 83.00 7 kswapd0 08:36:46 0 93 0.00 70.00 0.00 0.00 70.00 7 kswapd0 08:36:47 0 93 0.00 78.00 0.00 1.00 78.00 2 kswapd0 08:36:48 0 93 0.00 81.00 0.00 0.00 81.00 3 kswapd0 08:36:49 0 93 0.00 54.00 0.00 0.00 54.00 7 kswapd0 08:36:50 0 93 0.00 76.00 0.00 0.00 76.00 1 kswapd0 08:36:51 0 93 0.00 75.00 0.00 0.00 75.00 0 kswapd0 08:36:52 0 93 0.00 73.00 0.00 0.00 73.00 7 kswapd0 08:36:53 0 93 0.00 61.00 0.00 1.00 61.00 7 kswapd0 08:36:54 0 93 0.00 80.00 0.00 0.00 80.00 7 kswapd0 08:36:55 0 93 0.00 64.00 0.00 0.00 64.00 7 kswapd0 08:36:56 0 93 0.00 56.00 0.00 0.00 56.00 7 kswapd0 08:36:57 0 93 0.00 26.00 0.00 0.00 26.00 2 kswapd0 08:36:58 0 93 0.00 24.00 0.00 1.00 24.00 3 kswapd0 08:36:59 0 93 0.00 22.00 0.00 1.00 22.00 3 kswapd0 08:37:00 0 93 0.00 15.84 0.00 0.00 15.84 3 kswapd0 08:37:01 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 08:37:02 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 Read test data after using uncached buffer I/O: reading bs 32768, uncached 1 1s: 1863MB/sec, MB=1863 2s: 1903MB/sec, MB=3766 3s: 1860MB/sec, MB=5627 4s: 1864MB/sec, MB=7491 5s: 1860MB/sec, MB=9352 6s: 1854MB/sec, MB=11206 7s: 1874MB/sec, MB=13081 8s: 1874MB/sec, MB=14943 9s: 1840MB/sec, MB=16798 10s: 1849MB/sec, MB=18647 11s: 1863MB/sec, MB=20511 12s: 1798MB/sec, MB=22310 13s: 1897MB/sec, MB=24207 14s: 1817MB/sec, MB=26025 15s: 1893MB/sec, MB=27918 16s: 1917MB/sec, MB=29836 17s: 1863MB/sec, MB=31699 18s: 1904MB/sec, MB=33604 19s: 1894MB/sec, MB=35499 20s: 1907MB/sec, MB=37407 08:38:00 UID PID %usr %system %guest %wait %CPU CPU Command 08:38:01 0 93 0.00 0.00 0.00 0.00 0.00 4 kswapd0 08:38:02 0 93 0.00 0.00 0.00 0.00 0.00 4 kswapd0 08:38:03 0 93 0.00 0.00 0.00 0.00 0.00 4 kswapd0 08:38:04 0 93 0.00 0.00 0.00 0.00 0.00 4 kswapd0 08:38:05 0 93 0.00 0.00 0.00 0.00 0.00 4 kswapd0 08:38:06 0 93 0.00 1.00 0.00 1.00 1.00 0 kswapd0 08:38:07 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:08 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:09 0 93 0.00 1.00 0.00 0.00 1.00 1 kswapd0 08:38:10 0 93 0.00 0.00 0.00 0.00 0.00 1 kswapd0 08:38:11 0 93 0.00 0.00 0.00 0.00 0.00 1 kswapd0 08:38:12 0 93 0.00 0.00 0.00 0.00 0.00 1 kswapd0 08:38:13 0 93 0.00 0.00 0.00 0.00 0.00 1 kswapd0 08:38:14 0 93 0.00 0.00 0.00 0.00 0.00 1 kswapd0 08:38:15 0 93 0.00 3.00 0.00 0.00 3.00 0 kswapd0 08:38:16 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:17 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:18 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:19 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:20 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:21 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:22 0 93 0.00 0.00 0.00 0.00 0.00 0 kswapd0 08:38:23 0 93 0.00 3.00 0.00 0.00 3.00 4 kswapd0 08:38:24 0 93 0.00 0.00 0.00 0.00 0.00 4 kswapd0 08:38:25 0 93 0.00 0.00 0.00 0.00 0.00 4 kswapd0 08:38:26 0 93 0.00 4.00 0.00 0.00 4.00 3 kswapd0 08:38:27 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 08:38:28 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 08:38:29 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 08:38:30 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 08:38:31 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 08:38:32 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 08:38:33 0 93 0.00 0.00 0.00 0.00 0.00 3 kswapd0 [1] https://pastebin.com/u8eCBzB5 [2] https://lore.kernel.org/all/20241220154831.1086649-10-axboe@kernel.dk/T/#m0dff9e4f79c95a75c6b2cf202bc9d3d6f4559723 Thanks, Qi >> [1] >> https://lore.kernel.org/all/20241220154831.1086649-10-axboe@kernel.dk/T/#m58520a94b46f543d82db3711453dfc7bb594b2b0 >> >> Signed-off-by: Qi Han <hanqi@vivo.com> >> --- >> fs/f2fs/file.c | 2 +- >> 1 file changed, 1 insertion(+), 1 deletion(-) >> >> diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c >> index 696131e655ed..d8da1fc2febf 100644 >> --- a/fs/f2fs/file.c >> +++ b/fs/f2fs/file.c >> @@ -5425,5 +5425,5 @@ const struct file_operations f2fs_file_operations = { >> .splice_read = f2fs_file_splice_read, >> .splice_write = iter_file_splice_write, >> .fadvise = f2fs_file_fadvise, >> - .fop_flags = FOP_BUFFER_RASYNC, >> + .fop_flags = FOP_BUFFER_RASYNC | FOP_DONTCACHE, >> }; _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-15 3:10 [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O Qi Han via Linux-f2fs-devel 2025-07-15 6:58 ` Chao Yu via Linux-f2fs-devel @ 2025-07-15 14:28 ` Jens Axboe 2025-07-16 3:34 ` hanqi via Linux-f2fs-devel 1 sibling, 1 reply; 10+ messages in thread From: Jens Axboe @ 2025-07-15 14:28 UTC (permalink / raw) To: Qi Han, jaegeuk, chao; +Cc: linux-kernel, linux-f2fs-devel On 7/14/25 9:10 PM, Qi Han wrote: > Jens has already completed the development of uncached buffered I/O > in git [1], and in f2fs, the feature can be enabled simply by setting > the FOP_DONTCACHE flag in f2fs_file_operations. You need to ensure that for any DONTCACHE IO that the completion is routed via non-irq context, if applicable. I didn't verify that this is the case for f2fs. Generally you can deduce this as well through testing, I'd say the following cases would be interesting to test: 1) Normal DONTCACHE buffered read 2) Overwrite DONTCACHE buffered write 3) Append DONTCACHE buffered write Test those with DEBUG_ATOMIC_SLEEP set in your config, and it that doesn't complain, that's a great start. For the above test cases as well, verify that page cache doesn't grow as IO is performed. A bit is fine for things like meta data, but generally you want to see it remain basically flat in terms of page cache usage. Maybe this is all fine, like I said I didn't verify. Just mentioning it for completeness sake. -- Jens Axboe _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-15 14:28 ` Jens Axboe @ 2025-07-16 3:34 ` hanqi via Linux-f2fs-devel 2025-07-16 3:43 ` Jens Axboe 0 siblings, 1 reply; 10+ messages in thread From: hanqi via Linux-f2fs-devel @ 2025-07-16 3:34 UTC (permalink / raw) To: Jens Axboe, jaegeuk, chao; +Cc: linux-kernel, linux-f2fs-devel 在 2025/7/15 22:28, Jens Axboe 写道: > On 7/14/25 9:10 PM, Qi Han wrote: >> Jens has already completed the development of uncached buffered I/O >> in git [1], and in f2fs, the feature can be enabled simply by setting >> the FOP_DONTCACHE flag in f2fs_file_operations. > You need to ensure that for any DONTCACHE IO that the completion is > routed via non-irq context, if applicable. I didn't verify that this is > the case for f2fs. Generally you can deduce this as well through > testing, I'd say the following cases would be interesting to test: > > 1) Normal DONTCACHE buffered read > 2) Overwrite DONTCACHE buffered write > 3) Append DONTCACHE buffered write > > Test those with DEBUG_ATOMIC_SLEEP set in your config, and it that > doesn't complain, that's a great start. > > For the above test cases as well, verify that page cache doesn't grow as > IO is performed. A bit is fine for things like meta data, but generally > you want to see it remain basically flat in terms of page cache usage. > > Maybe this is all fine, like I said I didn't verify. Just mentioning it > for completeness sake. Hi, Jens Thanks for your suggestion. As I mentioned earlier in [1], in f2fs, the regular buffered write path invokes folio_end_writeback from a softirq context. Therefore, it seems that f2fs may not be suitable for DONTCACHE I/O writes. I’d like to ask a question: why is DONTCACHE I/O write restricted to non-interrupt context only? Is it because dropping the page might be too time-consuming to be done safely in interrupt context? This might be a naive question, but I’d really appreciate your clarification. Thanks in advance. [1] https://lore.kernel.org/all/137c0a07-ea0a-48fa-acc4-3e0ec63681f4@vivo.com/ > _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-16 3:34 ` hanqi via Linux-f2fs-devel @ 2025-07-16 3:43 ` Jens Axboe 2025-07-16 8:27 ` hanqi via Linux-f2fs-devel 0 siblings, 1 reply; 10+ messages in thread From: Jens Axboe @ 2025-07-16 3:43 UTC (permalink / raw) To: hanqi, jaegeuk, chao; +Cc: linux-kernel, linux-f2fs-devel On 7/15/25 9:34 PM, hanqi wrote: > > > ? 2025/7/15 22:28, Jens Axboe ??: >> On 7/14/25 9:10 PM, Qi Han wrote: >>> Jens has already completed the development of uncached buffered I/O >>> in git [1], and in f2fs, the feature can be enabled simply by setting >>> the FOP_DONTCACHE flag in f2fs_file_operations. >> You need to ensure that for any DONTCACHE IO that the completion is >> routed via non-irq context, if applicable. I didn't verify that this is >> the case for f2fs. Generally you can deduce this as well through >> testing, I'd say the following cases would be interesting to test: >> >> 1) Normal DONTCACHE buffered read >> 2) Overwrite DONTCACHE buffered write >> 3) Append DONTCACHE buffered write >> >> Test those with DEBUG_ATOMIC_SLEEP set in your config, and it that >> doesn't complain, that's a great start. >> >> For the above test cases as well, verify that page cache doesn't grow as >> IO is performed. A bit is fine for things like meta data, but generally >> you want to see it remain basically flat in terms of page cache usage. >> >> Maybe this is all fine, like I said I didn't verify. Just mentioning it >> for completeness sake. > > Hi, Jens > Thanks for your suggestion. As I mentioned earlier in [1], in f2fs, > the regular buffered write path invokes folio_end_writeback from a > softirq context. Therefore, it seems that f2fs may not be suitable > for DONTCACHE I/O writes. > > I?d like to ask a question: why is DONTCACHE I/O write restricted to > non-interrupt context only? Is it because dropping the page might be > too time-consuming to be done safely in interrupt context? This might > be a naive question, but I?d really appreciate your clarification. > Thanks in advance. Because (as of right now, at least) the code doing the invalidation needs process context. There are various reasons for this, which you'll see if you follow the path off folio_end_writeback() -> filemap_end_dropbehind_write() -> filemap_end_dropbehind() -> folio_unmap_invalidate(). unmap_mapping_folio() is one case, and while that may be doable, the inode i_lock is not IRQ safe. Most file systems have a need to punt some writeback completions to non-irq context, eg for file extending etc. Hence for most file systems, the dontcache case just becomes another case that needs to go through that path. It'd certainly be possible to improve upon this, for example by having an opportunistic dontcache unmap from IRQ/soft-irq context, and then punting to a workqueue if that doesn't pan out. But this doesn't exist as of yet, hence the need for the workqueue punt. -- Jens Axboe _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-16 3:43 ` Jens Axboe @ 2025-07-16 8:27 ` hanqi via Linux-f2fs-devel 2025-07-24 13:09 ` Chao Yu via Linux-f2fs-devel 0 siblings, 1 reply; 10+ messages in thread From: hanqi via Linux-f2fs-devel @ 2025-07-16 8:27 UTC (permalink / raw) To: Jens Axboe, jaegeuk, chao; +Cc: linux-kernel, linux-f2fs-devel 在 2025/7/16 11:43, Jens Axboe 写道: > On 7/15/25 9:34 PM, hanqi wrote: >> >> ? 2025/7/15 22:28, Jens Axboe ??: >>> On 7/14/25 9:10 PM, Qi Han wrote: >>>> Jens has already completed the development of uncached buffered I/O >>>> in git [1], and in f2fs, the feature can be enabled simply by setting >>>> the FOP_DONTCACHE flag in f2fs_file_operations. >>> You need to ensure that for any DONTCACHE IO that the completion is >>> routed via non-irq context, if applicable. I didn't verify that this is >>> the case for f2fs. Generally you can deduce this as well through >>> testing, I'd say the following cases would be interesting to test: >>> >>> 1) Normal DONTCACHE buffered read >>> 2) Overwrite DONTCACHE buffered write >>> 3) Append DONTCACHE buffered write >>> >>> Test those with DEBUG_ATOMIC_SLEEP set in your config, and it that >>> doesn't complain, that's a great start. >>> >>> For the above test cases as well, verify that page cache doesn't grow as >>> IO is performed. A bit is fine for things like meta data, but generally >>> you want to see it remain basically flat in terms of page cache usage. >>> >>> Maybe this is all fine, like I said I didn't verify. Just mentioning it >>> for completeness sake. >> Hi, Jens >> Thanks for your suggestion. As I mentioned earlier in [1], in f2fs, >> the regular buffered write path invokes folio_end_writeback from a >> softirq context. Therefore, it seems that f2fs may not be suitable >> for DONTCACHE I/O writes. >> >> I?d like to ask a question: why is DONTCACHE I/O write restricted to >> non-interrupt context only? Is it because dropping the page might be >> too time-consuming to be done safely in interrupt context? This might >> be a naive question, but I?d really appreciate your clarification. >> Thanks in advance. > Because (as of right now, at least) the code doing the invalidation > needs process context. There are various reasons for this, which you'll > see if you follow the path off folio_end_writeback() -> > filemap_end_dropbehind_write() -> filemap_end_dropbehind() -> > folio_unmap_invalidate(). unmap_mapping_folio() is one case, and while > that may be doable, the inode i_lock is not IRQ safe. > > Most file systems have a need to punt some writeback completions to > non-irq context, eg for file extending etc. Hence for most file systems, > the dontcache case just becomes another case that needs to go through > that path. > > It'd certainly be possible to improve upon this, for example by having > an opportunistic dontcache unmap from IRQ/soft-irq context, and then > punting to a workqueue if that doesn't pan out. But this doesn't exist > as of yet, hence the need for the workqueue punt. Hi, Jens Thank you for your response. I tested uncached buffer I/O reads with a 50GB dataset on a local F2FS filesystem, and the page cache size only increased slightly, which I believe aligns with expectations. After clearing the page cache, the page cache size returned to its initial state. The test results are as follows: stat 50G.txt File: 50G.txt Size: 53687091200 Blocks: 104960712 IO Blocks: 512 regular file [read before]: echo 3 > /proc/sys/vm/drop_caches 01:48:17 kbmemfree kbavail kbmemused %memused kbbuffers kbcached kbcommit %commit kbactive kbinact kbdirty 01:50:59 6404648 8149508 2719384 23.40 512 1898092 199384760 823.75 1846756 466832 44 ./uncached_io_test 8192 1 1 50G.txt Starting 1 threads reading bs 8192, uncached 1 1s: 754MB/sec, MB=754 ... 64s: 844MB/sec, MB=262144 [read after]: 01:52:33 6326664 8121240 2747968 23.65 728 1947656 199384788 823.75 1887896 502004 68 echo 3 > /proc/sys/vm/drop_caches 01:53:11 6351136 8096936 2772400 23.86 512 1900500 199385216 823.75 1847252 533768 104 Hi Chao, Given that F2FS currently calls folio_end_writeback in the softirq context for normal write scenarios, could we first support uncached buffer I/O reads? For normal uncached buffer I/O writes, would it be feasible for F2FS to introduce an asynchronous workqueue to handle the page drop operation in the future? What are your thoughts on this? Thank you! _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-16 8:27 ` hanqi via Linux-f2fs-devel @ 2025-07-24 13:09 ` Chao Yu via Linux-f2fs-devel 2025-07-25 1:44 ` hanqi via Linux-f2fs-devel 0 siblings, 1 reply; 10+ messages in thread From: Chao Yu via Linux-f2fs-devel @ 2025-07-24 13:09 UTC (permalink / raw) To: hanqi, Jens Axboe, jaegeuk; +Cc: linux-kernel, linux-f2fs-devel On 2025/7/16 16:27, hanqi wrote: > > > 在 2025/7/16 11:43, Jens Axboe 写道: >> On 7/15/25 9:34 PM, hanqi wrote: >>> >>> ? 2025/7/15 22:28, Jens Axboe ??: >>>> On 7/14/25 9:10 PM, Qi Han wrote: >>>>> Jens has already completed the development of uncached buffered I/O >>>>> in git [1], and in f2fs, the feature can be enabled simply by setting >>>>> the FOP_DONTCACHE flag in f2fs_file_operations. >>>> You need to ensure that for any DONTCACHE IO that the completion is >>>> routed via non-irq context, if applicable. I didn't verify that this is >>>> the case for f2fs. Generally you can deduce this as well through >>>> testing, I'd say the following cases would be interesting to test: >>>> >>>> 1) Normal DONTCACHE buffered read >>>> 2) Overwrite DONTCACHE buffered write >>>> 3) Append DONTCACHE buffered write >>>> >>>> Test those with DEBUG_ATOMIC_SLEEP set in your config, and it that >>>> doesn't complain, that's a great start. >>>> >>>> For the above test cases as well, verify that page cache doesn't grow as >>>> IO is performed. A bit is fine for things like meta data, but generally >>>> you want to see it remain basically flat in terms of page cache usage. >>>> >>>> Maybe this is all fine, like I said I didn't verify. Just mentioning it >>>> for completeness sake. >>> Hi, Jens >>> Thanks for your suggestion. As I mentioned earlier in [1], in f2fs, >>> the regular buffered write path invokes folio_end_writeback from a >>> softirq context. Therefore, it seems that f2fs may not be suitable >>> for DONTCACHE I/O writes. >>> >>> I?d like to ask a question: why is DONTCACHE I/O write restricted to >>> non-interrupt context only? Is it because dropping the page might be >>> too time-consuming to be done safely in interrupt context? This might >>> be a naive question, but I?d really appreciate your clarification. >>> Thanks in advance. >> Because (as of right now, at least) the code doing the invalidation >> needs process context. There are various reasons for this, which you'll >> see if you follow the path off folio_end_writeback() -> >> filemap_end_dropbehind_write() -> filemap_end_dropbehind() -> >> folio_unmap_invalidate(). unmap_mapping_folio() is one case, and while >> that may be doable, the inode i_lock is not IRQ safe. >> >> Most file systems have a need to punt some writeback completions to >> non-irq context, eg for file extending etc. Hence for most file systems, >> the dontcache case just becomes another case that needs to go through >> that path. >> >> It'd certainly be possible to improve upon this, for example by having >> an opportunistic dontcache unmap from IRQ/soft-irq context, and then >> punting to a workqueue if that doesn't pan out. But this doesn't exist >> as of yet, hence the need for the workqueue punt. Thanks Jens for the detailed explanation. > > Hi, Jens > Thank you for your response. I tested uncached buffer I/O reads with > a 50GB dataset on a local F2FS filesystem, and the page cache size > only increased slightly, which I believe aligns with expectations. > After clearing the page cache, the page cache size returned to its > initial state. The test results are as follows: > > stat 50G.txt > File: 50G.txt > Size: 53687091200 Blocks: 104960712 IO Blocks: 512 regular file > > [read before]: > echo 3 > /proc/sys/vm/drop_caches > 01:48:17 kbmemfree kbavail kbmemused %memused kbbuffers kbcached kbcommit %commit kbactive kbinact kbdirty > 01:50:59 6404648 8149508 2719384 23.40 512 1898092 199384760 823.75 1846756 466832 44 > > ./uncached_io_test 8192 1 1 50G.txt > Starting 1 threads > reading bs 8192, uncached 1 > 1s: 754MB/sec, MB=754 > ... > 64s: 844MB/sec, MB=262144 > > [read after]: > 01:52:33 6326664 8121240 2747968 23.65 728 1947656 199384788 823.75 1887896 502004 68 > echo 3 > /proc/sys/vm/drop_caches > 01:53:11 6351136 8096936 2772400 23.86 512 1900500 199385216 823.75 1847252 533768 104 > > Hi Chao, > Given that F2FS currently calls folio_end_writeback in the softirq > context for normal write scenarios, could we first support uncached > buffer I/O reads? For normal uncached buffer I/O writes, would it be > feasible for F2FS to introduce an asynchronous workqueue to handle the > page drop operation in the future? What are your thoughts on this? Qi, Sorry for the delay. I think it will be good to support uncached buffered I/O in read path first, and then let's take a look what we can do for write path, anyway, let's do this step by step. Can you please update the patch? - support read path only - include test data in commit message > Thank you! > > _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-24 13:09 ` Chao Yu via Linux-f2fs-devel @ 2025-07-25 1:44 ` hanqi via Linux-f2fs-devel 2025-07-25 2:37 ` Chao Yu via Linux-f2fs-devel 0 siblings, 1 reply; 10+ messages in thread From: hanqi via Linux-f2fs-devel @ 2025-07-25 1:44 UTC (permalink / raw) To: Chao Yu, Jens Axboe, jaegeuk; +Cc: linux-kernel, linux-f2fs-devel 在 2025/7/24 21:09, Chao Yu 写道: > On 2025/7/16 16:27, hanqi wrote: >> >> >> 在 2025/7/16 11:43, Jens Axboe 写道: >>> On 7/15/25 9:34 PM, hanqi wrote: >>>> >>>> ? 2025/7/15 22:28, Jens Axboe ??: >>>>> On 7/14/25 9:10 PM, Qi Han wrote: >>>>>> Jens has already completed the development of uncached buffered I/O >>>>>> in git [1], and in f2fs, the feature can be enabled simply by >>>>>> setting >>>>>> the FOP_DONTCACHE flag in f2fs_file_operations. >>>>> You need to ensure that for any DONTCACHE IO that the completion is >>>>> routed via non-irq context, if applicable. I didn't verify that >>>>> this is >>>>> the case for f2fs. Generally you can deduce this as well through >>>>> testing, I'd say the following cases would be interesting to test: >>>>> >>>>> 1) Normal DONTCACHE buffered read >>>>> 2) Overwrite DONTCACHE buffered write >>>>> 3) Append DONTCACHE buffered write >>>>> >>>>> Test those with DEBUG_ATOMIC_SLEEP set in your config, and it that >>>>> doesn't complain, that's a great start. >>>>> >>>>> For the above test cases as well, verify that page cache doesn't >>>>> grow as >>>>> IO is performed. A bit is fine for things like meta data, but >>>>> generally >>>>> you want to see it remain basically flat in terms of page cache >>>>> usage. >>>>> >>>>> Maybe this is all fine, like I said I didn't verify. Just >>>>> mentioning it >>>>> for completeness sake. >>>> Hi, Jens >>>> Thanks for your suggestion. As I mentioned earlier in [1], in f2fs, >>>> the regular buffered write path invokes folio_end_writeback from a >>>> softirq context. Therefore, it seems that f2fs may not be suitable >>>> for DONTCACHE I/O writes. >>>> >>>> I?d like to ask a question: why is DONTCACHE I/O write restricted to >>>> non-interrupt context only? Is it because dropping the page might be >>>> too time-consuming to be done safely in interrupt context? This might >>>> be a naive question, but I?d really appreciate your clarification. >>>> Thanks in advance. >>> Because (as of right now, at least) the code doing the invalidation >>> needs process context. There are various reasons for this, which you'll >>> see if you follow the path off folio_end_writeback() -> >>> filemap_end_dropbehind_write() -> filemap_end_dropbehind() -> >>> folio_unmap_invalidate(). unmap_mapping_folio() is one case, and while >>> that may be doable, the inode i_lock is not IRQ safe. >>> >>> Most file systems have a need to punt some writeback completions to >>> non-irq context, eg for file extending etc. Hence for most file >>> systems, >>> the dontcache case just becomes another case that needs to go through >>> that path. >>> >>> It'd certainly be possible to improve upon this, for example by having >>> an opportunistic dontcache unmap from IRQ/soft-irq context, and then >>> punting to a workqueue if that doesn't pan out. But this doesn't exist >>> as of yet, hence the need for the workqueue punt. > > Thanks Jens for the detailed explanation. > >> >> Hi, Jens >> Thank you for your response. I tested uncached buffer I/O reads with >> a 50GB dataset on a local F2FS filesystem, and the page cache size >> only increased slightly, which I believe aligns with expectations. >> After clearing the page cache, the page cache size returned to its >> initial state. The test results are as follows: >> >> stat 50G.txt >> File: 50G.txt >> Size: 53687091200 Blocks: 104960712 IO Blocks: 512 >> regular file >> >> [read before]: >> echo 3 > /proc/sys/vm/drop_caches >> 01:48:17 kbmemfree kbavail kbmemused %memused >> kbbuffers kbcached kbcommit %commit kbactive kbinact kbdirty >> 01:50:59 6404648 8149508 2719384 23.40 512 1898092 >> 199384760 823.75 1846756 466832 44 >> >> ./uncached_io_test 8192 1 1 50G.txt >> Starting 1 threads >> reading bs 8192, uncached 1 >> 1s: 754MB/sec, MB=754 >> ... >> 64s: 844MB/sec, MB=262144 >> >> [read after]: >> 01:52:33 6326664 8121240 2747968 23.65 728 1947656 >> 199384788 823.75 1887896 502004 68 >> echo 3 > /proc/sys/vm/drop_caches >> 01:53:11 6351136 8096936 2772400 23.86 512 1900500 >> 199385216 823.75 1847252 533768 104 >> >> Hi Chao, >> Given that F2FS currently calls folio_end_writeback in the softirq >> context for normal write scenarios, could we first support uncached >> buffer I/O reads? For normal uncached buffer I/O writes, would it be >> feasible for F2FS to introduce an asynchronous workqueue to handle the >> page drop operation in the future? What are your thoughts on this? > > Qi, > > Sorry for the delay. > > I think it will be good to support uncached buffered I/O in read path > first, and then let's take a look what we can do for write path, anyway, > let's do this step by step. > > Can you please update the patch? > - support read path only > - include test data in commit message Chao I will re-submit a patch to first enable F2FS support for uncached buffer I/O reads. Following that, I will work on implementing asynchronous page dropping in F2FS. Thank you! > >> Thank you! >> >> _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O 2025-07-25 1:44 ` hanqi via Linux-f2fs-devel @ 2025-07-25 2:37 ` Chao Yu via Linux-f2fs-devel 0 siblings, 0 replies; 10+ messages in thread From: Chao Yu via Linux-f2fs-devel @ 2025-07-25 2:37 UTC (permalink / raw) To: hanqi, Jens Axboe, jaegeuk; +Cc: linux-kernel, linux-f2fs-devel On 7/25/2025 9:44 AM, hanqi wrote: > > > 在 2025/7/24 21:09, Chao Yu 写道: >> On 2025/7/16 16:27, hanqi wrote: >>> >>> >>> 在 2025/7/16 11:43, Jens Axboe 写道: >>>> On 7/15/25 9:34 PM, hanqi wrote: >>>>> >>>>> ? 2025/7/15 22:28, Jens Axboe ??: >>>>>> On 7/14/25 9:10 PM, Qi Han wrote: >>>>>>> Jens has already completed the development of uncached buffered I/O >>>>>>> in git [1], and in f2fs, the feature can be enabled simply by >>>>>>> setting >>>>>>> the FOP_DONTCACHE flag in f2fs_file_operations. >>>>>> You need to ensure that for any DONTCACHE IO that the completion is >>>>>> routed via non-irq context, if applicable. I didn't verify that >>>>>> this is >>>>>> the case for f2fs. Generally you can deduce this as well through >>>>>> testing, I'd say the following cases would be interesting to test: >>>>>> >>>>>> 1) Normal DONTCACHE buffered read >>>>>> 2) Overwrite DONTCACHE buffered write >>>>>> 3) Append DONTCACHE buffered write >>>>>> >>>>>> Test those with DEBUG_ATOMIC_SLEEP set in your config, and it that >>>>>> doesn't complain, that's a great start. >>>>>> >>>>>> For the above test cases as well, verify that page cache doesn't >>>>>> grow as >>>>>> IO is performed. A bit is fine for things like meta data, but >>>>>> generally >>>>>> you want to see it remain basically flat in terms of page cache >>>>>> usage. >>>>>> >>>>>> Maybe this is all fine, like I said I didn't verify. Just >>>>>> mentioning it >>>>>> for completeness sake. >>>>> Hi, Jens >>>>> Thanks for your suggestion. As I mentioned earlier in [1], in f2fs, >>>>> the regular buffered write path invokes folio_end_writeback from a >>>>> softirq context. Therefore, it seems that f2fs may not be suitable >>>>> for DONTCACHE I/O writes. >>>>> >>>>> I?d like to ask a question: why is DONTCACHE I/O write restricted to >>>>> non-interrupt context only? Is it because dropping the page might be >>>>> too time-consuming to be done safely in interrupt context? This might >>>>> be a naive question, but I?d really appreciate your clarification. >>>>> Thanks in advance. >>>> Because (as of right now, at least) the code doing the invalidation >>>> needs process context. There are various reasons for this, which you'll >>>> see if you follow the path off folio_end_writeback() -> >>>> filemap_end_dropbehind_write() -> filemap_end_dropbehind() -> >>>> folio_unmap_invalidate(). unmap_mapping_folio() is one case, and while >>>> that may be doable, the inode i_lock is not IRQ safe. >>>> >>>> Most file systems have a need to punt some writeback completions to >>>> non-irq context, eg for file extending etc. Hence for most file >>>> systems, >>>> the dontcache case just becomes another case that needs to go through >>>> that path. >>>> >>>> It'd certainly be possible to improve upon this, for example by having >>>> an opportunistic dontcache unmap from IRQ/soft-irq context, and then >>>> punting to a workqueue if that doesn't pan out. But this doesn't exist >>>> as of yet, hence the need for the workqueue punt. >> >> Thanks Jens for the detailed explanation. >> >>> >>> Hi, Jens >>> Thank you for your response. I tested uncached buffer I/O reads with >>> a 50GB dataset on a local F2FS filesystem, and the page cache size >>> only increased slightly, which I believe aligns with expectations. >>> After clearing the page cache, the page cache size returned to its >>> initial state. The test results are as follows: >>> >>> stat 50G.txt >>> File: 50G.txt >>> Size: 53687091200 Blocks: 104960712 IO Blocks: 512 >>> regular file >>> >>> [read before]: >>> echo 3 > /proc/sys/vm/drop_caches >>> 01:48:17 kbmemfree kbavail kbmemused %memused >>> kbbuffers kbcached kbcommit %commit kbactive kbinact kbdirty >>> 01:50:59 6404648 8149508 2719384 23.40 512 1898092 >>> 199384760 823.75 1846756 466832 44 >>> >>> ./uncached_io_test 8192 1 1 50G.txt >>> Starting 1 threads >>> reading bs 8192, uncached 1 >>> 1s: 754MB/sec, MB=754 >>> ... >>> 64s: 844MB/sec, MB=262144 >>> >>> [read after]: >>> 01:52:33 6326664 8121240 2747968 23.65 728 1947656 >>> 199384788 823.75 1887896 502004 68 >>> echo 3 > /proc/sys/vm/drop_caches >>> 01:53:11 6351136 8096936 2772400 23.86 512 1900500 >>> 199385216 823.75 1847252 533768 104 >>> >>> Hi Chao, >>> Given that F2FS currently calls folio_end_writeback in the softirq >>> context for normal write scenarios, could we first support uncached >>> buffer I/O reads? For normal uncached buffer I/O writes, would it be >>> feasible for F2FS to introduce an asynchronous workqueue to handle the >>> page drop operation in the future? What are your thoughts on this? >> >> Qi, >> >> Sorry for the delay. >> >> I think it will be good to support uncached buffered I/O in read path >> first, and then let's take a look what we can do for write path, anyway, >> let's do this step by step. >> >> Can you please update the patch? >> - support read path only >> - include test data in commit message > Chao > > I will re-submit a patch to first enable F2FS support for uncached > buffer I/O reads. Following that, I will work on implementing > asynchronous page dropping in F2FS. Qi, sure, please go ahead, thanks for the work. :) Thanks, > > Thank you! >> >>> Thank you! >>> >>> > _______________________________________________ Linux-f2fs-devel mailing list Linux-f2fs-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel ^ permalink raw reply [flat|nested] 10+ messages in thread
end of thread, other threads:[~2025-07-25 2:37 UTC | newest] Thread overview: 10+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2025-07-15 3:10 [f2fs-dev] [PATCH] f2fs: f2fs supports uncached buffered I/O Qi Han via Linux-f2fs-devel 2025-07-15 6:58 ` Chao Yu via Linux-f2fs-devel 2025-07-15 8:14 ` hanqi via Linux-f2fs-devel 2025-07-15 14:28 ` Jens Axboe 2025-07-16 3:34 ` hanqi via Linux-f2fs-devel 2025-07-16 3:43 ` Jens Axboe 2025-07-16 8:27 ` hanqi via Linux-f2fs-devel 2025-07-24 13:09 ` Chao Yu via Linux-f2fs-devel 2025-07-25 1:44 ` hanqi via Linux-f2fs-devel 2025-07-25 2:37 ` Chao Yu via Linux-f2fs-devel
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).