linux-kernel - Re: [PATCH v3 1/3] mm: kmemleak: Make the tool tolerant to struct scan

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 3 Oct 2019 04:36:39 -0400
From:   Qian Cai <cai@....pw>
To:     Alexey Kardashevskiy <aik@...abs.ru>
Cc:     Catalin Marinas <catalin.marinas@....com>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        Michal Hocko <mhocko@...nel.org>,
        Matthew Wilcox <willy@...radead.org>
Subject: Re: [PATCH v3 1/3] mm: kmemleak: Make the tool tolerant to struct scan_area allocation failures



> On Oct 3, 2019, at 2:13 AM, Alexey Kardashevskiy <aik@...abs.ru> wrote:
> 
> I came across this one while bisecting sudden drop in throughput of a 100Gbit Mellanox CX4 ethernet card in a PPC POWER9
> system, the speed dropped from 100Gbit to about 40Gbit. Bisect pointed at dba82d943177, this are the relevant config
> options:
> 
> [fstn1-p1 kernel]$ grep KMEMLEAK ~/pbuild/kernel-le-4g/.config
> CONFIG_HAVE_DEBUG_KMEMLEAK=y
> CONFIG_DEBUG_KMEMLEAK=y
> CONFIG_DEBUG_KMEMLEAK_EARLY_LOG_SIZE=16000
> # CONFIG_DEBUG_KMEMLEAK_TEST is not set
> # CONFIG_DEBUG_KMEMLEAK_DEFAULT_OFF is not set
> CONFIG_DEBUG_KMEMLEAK_AUTO_SCAN=y
> 
> Setting CONFIG_DEBUG_KMEMLEAK_MEM_POOL_SIZE=400 or even 4000 (this is what KMEMLEAK_EARLY_LOG_SIZE is now in the master)
> produces soft lockups on the recent upstream (sha1 a3c0e7b1fe1f):
> 
> [c000001fde64fb60] [c000000000c24ed4] _raw_write_unlock_irqrestore+0x54/0x70
> [c000001fde64fb90] [c0000000004117e4] find_and_remove_object+0xa4/0xd0
> [c000001fde64fbe0] [c000000000411c74] delete_object_full+0x24/0x50
> [c000001fde64fc00] [c000000000411d28] __kmemleak_do_cleanup+0x88/0xd0
> [c000001fde64fc40] [c00000000012a1a4] process_one_work+0x374/0x6a0
> [c000001fde64fd20] [c00000000012a548] worker_thread+0x78/0x5a0
> [c000001fde64fdb0] [c000000000135508] kthread+0x198/0x1a0
> [c000001fde64fe20] [c00000000000b980] ret_from_kernel_thread+0x5c/0x7c
> 
> KMEMLEAK_EARLY_LOG_SIZE=8000 works but slow.
> 
> Interestingly KMEMLEAK_EARLY_LOG_SIZE=400 on dba82d943177 still worked and I saw my 100Gbit. Disabling KMEMLEAK also
> fixes the speed (apparently).
> 
> Is that something expected? Thanks,

It is expected that a debug option like this will make the system slower.