Message-ID: <CAJfuBxzGJvrJo9nTXxZ3xZ7QmdSb6YxBw-bojZjQTpACBeK_sQ@mail.gmail.com>
Date:   Tue, 4 Apr 2023 19:38:41 -0600
From:   jim.cromie@...il.com
To:     Luis Chamberlain <mcgrof@...nel.org>
Cc:     Song Liu <song@...nel.org>,
        Catalin Marinas <catalin.marinas@....com>,
        linux-modules@...r.kernel.org, LKML <linux-kernel@...r.kernel.org>,
        Jason Baron <jbaron@...mai.com>,
        Greg KH <gregkh@...uxfoundation.org>
Subject: Re: kmemleaks on ac3b43283923 ("module: replace module_layout with module_memory")

On Mon, Apr 3, 2023 at 2:44 PM Luis Chamberlain <mcgrof@...nel.org> wrote:
>
> On Fri, Mar 31, 2023 at 05:27:04PM -0700, Song Liu wrote:
> > On Fri, Mar 31, 2023 at 12:00 AM Luis Chamberlain <mcgrof@...nel.org> wrote:
> > >
> > > On Thu, Mar 30, 2023 at 04:45:43PM -0600, jim.cromie@...il.com wrote:
> > > > hi Luis, etal
> > > >
> > > > kmemleak is reporting 19 leaks during boot
> > > >
> > > > because the hexdumps appeared to have module-names,
> > > > and Ive been hacking nearby, and see the same names
> > > > every time I boot my test-vm, I needed a clearer picture
> > > > Jason corroborated and bisected.
> > > >
> > > > the 19 leaks split into 2 groups,
> > > > 9 with names of builtin modules in the hexdump,
> > > > all with the same backtrace
> > > > 9 without module-names (with a shared backtrace)
> > > > +1 wo name-ish and a separate backtrace
> > >
> > > Song, please take a look.
> >
> > I will look into this next week.
>
> I'm thinking this may be it, at least this gets us to what we used to do
> as per original Catalinas' 4f2294b6dc88d ("kmemleak: Add modules
> support") and right before Song's patch.
>
> diff --git a/kernel/module/main.c b/kernel/module/main.c
> index 6b6da80f363f..3b9c71fa6096 100644
> --- a/kernel/module/main.c
> +++ b/kernel/module/main.c
> @@ -2240,7 +2240,10 @@ static int move_module(struct module *mod, struct load_info *info)
>                  * which is inside the block. Just mark it as not being a
>                  * leak.
>                  */
> -               kmemleak_ignore(ptr);
> +               if (type == MOD_INIT_TEXT)
> +                       kmemleak_ignore(ptr);
> +               else
> +                       kmemleak_not_leak(ptr);
>                 if (!ptr) {
>                         t = type;
>                         goto out_enomem;
>
> We used to use the grey area for the TEXT but the original commit
> doesn't explain too well why we grey out init but not the others. Ie
> why kmemleak_ignore() on init and kmemleak_not_leak() on the others.
>
> Catalinas, any thoughts / suggestions? Should we just stick to
> kmemleak_not_leak() for both now?
>
>   Luis

So I have mixed results.

your patch fixed the 19 leaks on my worktree / branch where I found them.

on top of
ac3b43283923 module: replace module_layout with module_memory

it fixed the (same) 19, but gets a few new ones.
whats weird is that once they report, they disappear from
/sys/kernel/debug/kmemleak

heres that kmemleak report, with a little preceding / setup,
performed by this bash scripting

drms_unload() {
    for m in i915 amdgpu nouveau \
iommu_v2 video i2c_algo_bit mxm_wmi wmi intel_rapl_msr \
drm_display_helper cec drm_kms_helper drm_ttm_helper ttm gpu_sched
drm_buddy drm;
    do
rmmod $m ;
    done
}
drms_load() {
    uname -a
    for m in i915 amdgpu nouveau;
    do
       modprobe $m $*
    done
}
cycle_drms() {
    for i in 1 $*; do # loop 1+argc times
      time drms_load
      drms_unload
    done
}

leak_drive () {

    [[ -f /sys/kernel/debug/kmemleak ]] || {
       echo "need KMEMLEAK"
       return
    }
    let count=0
    #echo ok testing, each dot is 10 secs
    while true; do

var=`cat /sys/kernel/debug/kmemleak`
if [[ -z $var ]] ; then
    cycle_drms
    echo scan >/sys/kernel/debug/kmemleak
else
    break
fi
((count=$count+1))
echo finished pass $count
    done
    cat /sys/kernel/debug/kmemleak
    dmesg | grep /sys/kernel/debug/kmemleak
    uname -a
}


finished pass 6
Linux (none) 6.3.0-rc1-f2-00002-g30504a44c558 #360 SMP PREEMPT_DYNAMIC
Tue Apr  4 15:25:05 MDT 2023 x86_64 x86_64 x86_64 GNU/Linux
[   51.768797] ACPI: bus type drm_connector registered
[   52.039443] AMD-Vi: AMD IOMMUv2 functionality not available on this
system - This is not a bug.
[   52.795766] [drm] amdgpu kernel modesetting enabled.
[   52.796288] amdgpu: CRAT table not found
[   52.796502] amdgpu: Virtual CRAT table created for CPU
[   52.796964] amdgpu: Topology: Add CPU node

real 0m1.354s
user 0m0.002s
sys 0m0.919s
rmmod: ERROR: Module intel_rapl_msr is not currently loaded
[   53.401823] ACPI: bus type drm_connector unregistered
[   53.595705] kmemleak: 2 new suspected memory leaks (see
/sys/kernel/debug/kmemleak)
finished pass 7
unreferenced object 0xffff8880059b0240 (size 192):
  comm "modprobe", pid 716, jiffies 4294714739 (age 6.065s)
  hex dump (first 32 bytes):
    00 db 50 c0 ff ff ff ff 00 00 00 00 00 00 00 00  ..P.............
    00 00 00 00 00 00 00 00 ea ff ff ff ff ff ff ff  ................
  backtrace:
    [<00000000406104d4>] __kmalloc+0x49/0x150
    [<00000000fe00c883>] __register_sysctl_table+0x51/0x7f0
    [<00000000438011af>] 0xffffffffc04faa78
    [<000000009a44098c>] 0xffffffffc037f01b
    [<00000000de0b0c0b>] do_one_initcall+0x43/0x210
    [<0000000016200549>] do_init_module+0x60/0x240
    [<00000000e5f75cca>] __do_sys_finit_module+0x93/0xf0
    [<0000000014ed2961>] do_syscall_64+0x34/0x80
    [<00000000d14e8c97>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
unreferenced object 0xffff88800ce25800 (size 256):
  comm "modprobe", pid 716, jiffies 4294714739 (age 6.065s)
  hex dump (first 32 bytes):
    78 58 e2 0c 80 88 ff ff 00 00 00 00 00 00 00 00  xX..............
    00 00 00 00 00 00 00 00 ea ff ff ff ff ff ff ff  ................
  backtrace:
    [<00000000406104d4>] __kmalloc+0x49/0x150
    [<00000000ec6658c8>] __register_sysctl_table+0x569/0x7f0
    [<00000000438011af>] 0xffffffffc04faa78
    [<000000009a44098c>] 0xffffffffc037f01b
    [<00000000de0b0c0b>] do_one_initcall+0x43/0x210
    [<0000000016200549>] do_init_module+0x60/0x240
    [<00000000e5f75cca>] __do_sys_finit_module+0x93/0xf0
    [<0000000014ed2961>] do_syscall_64+0x34/0x80
    [<00000000d14e8c97>] entry_SYSCALL_64_after_hwframe+0x46/0xb0
[   53.595705] kmemleak: 2 new suspected memory leaks (see
/sys/kernel/debug/kmemleak)
Linux (none) 6.3.0-rc1-f2-00002-g30504a44c558 #360 SMP PREEMPT_DYNAMIC
Tue Apr  4 15:25:05 MDT 2023 x86_64 x86_64 x86_64 GNU/Linux

at this point, kmemleak is empty.
Im guessing thats because the leak was in / under  do_init_module,
and __init mem is recycled.

Maybe its also why the leak-trace has 2 entries without symbol info
Heres the levels above & below those mystery levels

(gdb) l *(do_one_initcall+0x43)
0xffffffff81001093 is in do_one_initcall
(/home/jimc/projects/lx/wk-next/init/main.c:1306).
1301
1302 if (initcall_blacklisted(fn))
1303 return -EPERM;
1304
1305 do_trace_initcall_start(fn);
1306 ret = fn();
1307 do_trace_initcall_finish(fn, ret);
1308
1309 msgbuf[0] = 0;
1310
(gdb) l *(__register_sysctl_table+0x569)
0xffffffff814e5e99 is in __register_sysctl_table
(/home/jimc/projects/lx/wk-next/fs/proc/proc_sysctl.c:974).
969 char *new_name;
970
971 new = kzalloc(sizeof(*new) + sizeof(struct ctl_node) +
972       sizeof(struct ctl_table)*2 +  namelen + 1,
973       GFP_KERNEL);
974 if (!new)
975 return NULL;
976
977 node = (struct ctl_node *)(new + 1);
978 table = (struct ctl_table *)(node + 1);
(gdb)

Powered by blists - more mailing lists