linux-kernel - Re: [PATCH] selftests/ipc: skip msgque test if MSG

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] [day] [month] [year] [list]

Message-ID: <CALAjK2ipX33hHDWZVhd_WUKYyD-JvCL67YgOeQDhPUuP2ovM1g@mail.gmail.com>
Date: Sat, 24 Jan 2026 17:35:53 +0900
From: 조우열 <jouyeol8739@...il.com>
To: Shuah <shuah@...nel.org>
Cc: linux-kselftest@...r.kernel.org, linux-kernel@...r.kernel.org, 
	richard.weiyang@...il.com, akpm@...ux-foundation.org, sef1548@...il.com
Subject: Re: [PATCH] selftests/ipc: skip msgque test if MSG_COPY is not supported

Hi Shuah,

Thank you for your feedback.

I have verified the behavior on kernel 6.19.0-rc1+ by testing both
kernel configurations (with and without CONFIG_CHECKPOINT_RESTORE).

Regarding the MSG_COPY handling in ipc/msg.c (between lines 1098-1262),
my analysis of do_msgrcv() shows the following execution flow:

1. Early Return (Lines 1112-1118):
   While the check 'if (msgflg & MSG_COPY)' is indeed outside the
   #ifdef blocks for validation, it immediately calls prepare_copy().
   Since prepare_copy() is stubbed to return ERR_PTR(-ENOSYS) when
   CONFIG_CHECKPOINT_RESTORE is disabled (line 1066), the system call
   fails and returns early at line 1118.

2. Unreachable Logic (Lines 1158-1161):
   The subsequent logic that calls copy_msg() is only reached if
   prepare_copy() succeeds. Without the config, this entire path
   becomes unreachable "dead code" in practice.

This confirms that while the flag handling is visible in the source,
the actual functionality is strictly dependent on the configuration.

To demonstrate the necessity of this patch, I ran the mainline test
in both environments:

1. With CONFIG_CHECKPOINT_RESTORE=y:
   # ./msgque
   # Totals: pass:0 fail:0 xfail:0 xpass:0 skip:0 error:0
   ok 1 selftests: ipc: msgque
   (Note: 'pass:0' is normal for this test as msgque.c does not
   explicitly call ksft_test_result_pass() for individual steps.)

2. With CONFIG_CHECKPOINT_RESTORE disabled (Mainline behavior):
   # ./msgque
   not ok 1 Failed to copy IPC message: Function not implemented (38)
   not ok 2 Failed to dump queue: -38
   # Totals: pass:0 fail:2 xfail:0 xpass:0 skip:0 error:0

Current mainline reports a "FAIL" for what is essentially a missing
kernel feature. My patch changes this to a "SKIP", which aligns
with kselftest standards.

With the patch applied, the result correctly becomes:
1..0 # SKIP MSG_COPY not supported

Best regards,
UYeol Jo