| Message ID | 20250410112117.2851-1-lirongqing@baidu.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | fs: Make file-nr output the total allocated file handles | expand |
On Thu, Apr 10, 2025 at 07:21:17PM +0800, lirongqing wrote: > From: Li RongQing <lirongqing@baidu.com> > > Make file-nr output the total allocated file handles, not per-cpu > cache number, it's more precise, and not in hot path > > Signed-off-by: Li RongQing <lirongqing@baidu.com> > --- That means grabbing a lock suddenly. Is there an actual use-case behind this? > fs/file_table.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/fs/file_table.c b/fs/file_table.c > index c04ed94..138114d 100644 > --- a/fs/file_table.c > +++ b/fs/file_table.c > @@ -102,7 +102,7 @@ EXPORT_SYMBOL_GPL(get_max_files); > static int proc_nr_files(const struct ctl_table *table, int write, void *buffer, > size_t *lenp, loff_t *ppos) > { > - files_stat.nr_files = get_nr_files(); > + files_stat.nr_files = percpu_counter_sum_positive(&nr_files); > return proc_doulongvec_minmax(table, write, buffer, lenp, ppos); > } > > -- > 2.9.4 >
On Fri, Apr 11, 2025 at 04:16:08PM +0200, Christian Brauner wrote: > On Thu, Apr 10, 2025 at 07:21:17PM +0800, lirongqing wrote: > > From: Li RongQing <lirongqing@baidu.com> > > > > Make file-nr output the total allocated file handles, not per-cpu > > cache number, it's more precise, and not in hot path > > > > Signed-off-by: Li RongQing <lirongqing@baidu.com> > > --- > > That means grabbing a lock suddenly. Is there an actual use-case > behind this? > The centralized value can be really grossly inaccurate as CPU count increases. There is some talks about fixing that, see: https://lore.kernel.org/linux-mm/20250410175149.1206995-1-mathieu.desnoyers@efficios.com/ In the meantime, given that this is only accessed when reading the /proc file, this should be fine? Note it still wont delay bumps/decs as long as they fit the batch (which is the common case).
> On Thu, Apr 10, 2025 at 07:21:17PM +0800, lirongqing wrote: > > From: Li RongQing <lirongqing@baidu.com> > > > > Make file-nr output the total allocated file handles, not per-cpu > > cache number, it's more precise, and not in hot path > > > > Signed-off-by: Li RongQing <lirongqing@baidu.com> > > --- > > That means grabbing a lock suddenly. Is there an actual use-case behind this? want to monitor it to find the leaking of the allocated file handler by some applications, but find it is inaccurate So I think this is not a hot path, percpu_counter_sum_positive may be able to be used Thank -Li
On Thu, 10 Apr 2025 19:21:17 +0800, lirongqing wrote: > Make file-nr output the total allocated file handles, not per-cpu > cache number, it's more precise, and not in hot path > > Applied to the vfs-6.16.misc branch of the vfs/vfs.git tree. Patches in the vfs-6.16.misc branch should appear in linux-next soon. Please report any outstanding bugs that were missed during review in a new review to the original patch series allowing us to drop it. It's encouraged to provide Acked-bys and Reviewed-bys even though the patch has now been applied. If possible patch trailers will be updated. Note that commit hashes shown below are subject to change due to rebase, trailer updates or similar. If in doubt, please check the listed branch. tree: https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git branch: vfs-6.16.misc [1/1] fs: Make file-nr output the total allocated file handles https://git.kernel.org/vfs/vfs/c/66319519f89d
> On Thu, 10 Apr 2025 19:21:17 +0800, lirongqing wrote: > > Make file-nr output the total allocated file handles, not per-cpu > > cache number, it's more precise, and not in hot path > > > > > Hi Christian Brauner: Could we change the changelog as below, it maybe more reasonable fs: Use percpu_counter_sum_positive in proc_nr_files The centralized value of percpu counter can be really grossly inaccurate as CPU count increases. and proc_nr_files is only accessed when reading proc file, not a hot path, so use percpu_counter_sum_positive in it, to make /proc/sys/fs/file-nr more accurate thanks -Li > Applied to the vfs-6.16.misc branch of the vfs/vfs.git tree. > Patches in the vfs-6.16.misc branch should appear in linux-next soon. > > Please report any outstanding bugs that were missed during review in a new > review to the original patch series allowing us to drop it. > > It's encouraged to provide Acked-bys and Reviewed-bys even though the patch > has now been applied. If possible patch trailers will be updated. > > Note that commit hashes shown below are subject to change due to rebase, > trailer updates or similar. If in doubt, please check the listed branch. > > tree: https://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs.git > branch: vfs-6.16.misc > > [1/1] fs: Make file-nr output the total allocated file handles > https://git.kernel.org/vfs/vfs/c/66319519f89d
diff --git a/fs/file_table.c b/fs/file_table.c index c04ed94..138114d 100644 --- a/fs/file_table.c +++ b/fs/file_table.c @@ -102,7 +102,7 @@ EXPORT_SYMBOL_GPL(get_max_files); static int proc_nr_files(const struct ctl_table *table, int write, void *buffer, size_t *lenp, loff_t *ppos) { - files_stat.nr_files = get_nr_files(); + files_stat.nr_files = percpu_counter_sum_positive(&nr_files); return proc_doulongvec_minmax(table, write, buffer, lenp, ppos); }