Re: [CFT] Re: VFS deadlock ?
[Posted April 2, 2013 by mkerrisk]
| From: |
| Al Viro <viro-AT-ZenIV.linux.org.uk> |
| To: |
| Linus Torvalds <torvalds-AT-linux-foundation.org> |
| Subject: |
| Re: [CFT] Re: VFS deadlock ? |
| Date: |
| Fri, 22 Mar 2013 06:22:32 +0000 |
| Message-ID: |
| <20130322062232.GO21522@ZenIV.linux.org.uk> |
| Cc: |
| Dave Jones <davej-AT-redhat.com>,
Linux Kernel <linux-kernel-AT-vger.kernel.org>,
"Eric W. Biederman" <ebiederm-AT-xmission.com> |
| Archive‑link: | |
Article |
On Fri, Mar 22, 2013 at 06:09:53AM +0000, Al Viro wrote:
> Hell knows... At the very least, I'd expect /proc/self to be fairly hot.
> During the boot time - /proc/mounts, /proc/filesystems, /proc/cmdline...
> Dunno. Would be interesting to slap a printk into proc_lookup_de() and
> see how much (and what) does it hit on a busy system...
FWIW, I'd done a quick and dirty stats collector right now - added hit count
into proc_dir_entry + /proc/counts walking the pde tree and dumping the
counts. Left it running in practically idle kvm testbox, after boot it
shows
kmsg => 2
kcore => 4
version => 4
uptime => 3
stat => 38
meminfo => 22
loadavg => 2
devices => 4
consoles => 4
cmdline => 109
filesystems => 118
swaps => 8
modules => 2
misc => 1
acpi => 6
fb => 1
counts => 14
sys => 1
bus => 1
fs => 1
net => 18
mounts => 10
self => 123
IOW, counts on this one are very low so far. OTOH, that kvm image is
practically idle, doesn't have desktop shite on it, etc. This really
ought to be checked on something busy... The first impression is that
the stuff outside of /proc/<pid> and /proc/sys isn't hot enough to care
about any cache retention...
