Resource leak or dead lock in the function task_dentry_path?

Resource leak or dead lock in the function task_dentry_path?

[Arkady]

Hi,

I have the following script

function task_cwd_full_path(pid)
{
        task = task_current()
        path = task_cwd_path(task)
        s = fullpath_struct_path(path)
        return s
}

global PATHS%;

probe kernel.function("do_execve")
{
        s = task_cwd_full_path(pid())
        PATHS[tid()] = s
}

probe syscall.fchdir.return
{
        s = task_cwd_full_path(pid())
        PATHS[tid()] = s
}

global Tick;

probe begin
{
        Tick = 0;
}

probe timer.sec(1)
{
        Tick = Tick + 1
        printf("Time=%4u\n", Tick)
}

I run the following loop in the shell

sudo find / -name '*' -type f -exec stat  {}  \;

I am getting "find: cannot fork: Resource temporarily unavailable" in
~60s on  ubuntu 4.4.0-59

What am I missing?

Thanks

Re: Resource leak or dead lock in the function task_dentry_path?

[Arkady]

Update.

Apparently task_cwd_path() is an evil. This modification also causes the failure

function task_cwd_full_path(pid)
{
        task = task_current()
        path = task_cwd_path(task)
        return path
}

On Thu, Jan 26, 2017 at 5:31 PM, Arkady <larytet@gmail.com> wrote:
> Hi,
>
> I have the following script
>
> function task_cwd_full_path(pid)
> {
>         task = task_current()
>         path = task_cwd_path(task)
>         s = fullpath_struct_path(path)
>         return s
> }
>
> global PATHS%;
>
> probe kernel.function("do_execve")
> {
>         s = task_cwd_full_path(pid())
>         PATHS[tid()] = s
> }
>
> probe syscall.fchdir.return
> {
>         s = task_cwd_full_path(pid())
>         PATHS[tid()] = s
> }
>
> global Tick;
>
> probe begin
> {
>         Tick = 0;
> }
>
> probe timer.sec(1)
> {
>         Tick = Tick + 1
>         printf("Time=%4u\n", Tick)
> }
>
> I run the following loop in the shell
>
> sudo find / -name '*' -type f -exec stat  {}  \;
>
> I am getting "find: cannot fork: Resource temporarily unavailable" in
> ~60s on  ubuntu 4.4.0-59
>
> What am I missing?
>
> Thanks

Re: Resource leak or dead lock in the function task_dentry_path?

[David Smith]

On 01/26/2017 11:11 AM, Arkady wrote:
> Update.
> 
> Apparently task_cwd_path() is an evil. This modification also causes the failure

That's <https://bugzilla.redhat.com/show_bug.cgi?id=1412691>, fixed by
the following commit:

====
commit e519829cd133be935ca9e3cc8e7251b0e74b20f4
Author: Kyle Walker <kwalker@redhat.com>
Date:   Thu Jan 12 10:11:39 2017 -0500

    RHBZ1412691 - fix kernel mem leak during tapset task_cwd_path()
====

-- 
David Smith
dsmith@redhat.com
Red Hat
http://www.redhat.com
256.217.0141 (direct)
256.837.0057 (fax)

Re: Resource leak or dead lock in the function task_dentry_path?

[Arkady]

Problem solved.
Great tip! Thanks!


On Thu, Jan 26, 2017 at 7:47 PM, David Smith <dsmith@redhat.com> wrote:
> On 01/26/2017 11:11 AM, Arkady wrote:
>> Update.
>>
>> Apparently task_cwd_path() is an evil. This modification also causes the failure
>
> That's <https://bugzilla.redhat.com/show_bug.cgi?id=1412691>, fixed by
> the following commit:
>
> ====
> commit e519829cd133be935ca9e3cc8e7251b0e74b20f4
> Author: Kyle Walker <kwalker@redhat.com>
> Date:   Thu Jan 12 10:11:39 2017 -0500
>
>     RHBZ1412691 - fix kernel mem leak during tapset task_cwd_path()
> ====
>
> --
> David Smith
> dsmith@redhat.com
> Red Hat
> http://www.redhat.com
> 256.217.0141 (direct)
> 256.837.0057 (fax)