linux-kernel - Re: [PATCH 1/2] kselftests: mm: Fix wrong __NR

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]

Message-ID: <e2a4d2b4-ca3f-4d21-82d5-b87bc9de9358@linuxfoundation.org>
Date: Fri, 20 Sep 2024 08:59:12 -0600
From: Shuah Khan <skhan@...uxfoundation.org>
To: Muhammad Usama Anjum <Usama.Anjum@...labora.com>,
 Andrew Morton <akpm@...ux-foundation.org>, Shuah Khan <shuah@...nel.org>,
 David Hildenbrand <david@...hat.com>, Peter Xu <peterx@...hat.com>
Cc: kernel@...labora.com, linux-mm@...ck.org,
 linux-kselftest@...r.kernel.org, linux-kernel@...r.kernel.org,
 John Hubbard <jhubbard@...dia.com>, Shuah Khan <skhan@...uxfoundation.org>
Subject: Re: [PATCH 1/2] kselftests: mm: Fix wrong __NR_userfaultfd value

On 9/17/24 23:46, Muhammad Usama Anjum wrote:
> On 9/17/24 6:56 AM, Shuah Khan wrote:
>> On 9/16/24 00:32, Muhammad Usama Anjum wrote:
>>> On 9/12/24 8:44 PM, Shuah Khan wrote:
>>>> On 9/12/24 04:31, Muhammad Usama Anjum wrote:
>>>>> The value of __NR_userfaultfd was changed to 282 when
>>>>> asm-generic/unistd.h was included. It makes the test to fail every time
>>>>> as the correct number of this syscall on x86_64 is 323. Fix the header
>>>>> to asm/unistd.h.
>>>>>
>>>>
>>>> "please elaborate every time" - I just built on my x86_64 and built
>>>> just fine.
>>> The build isn't broken.
>>>
>>>> I am not saying this isn't a problem, it is good to
>>>> understand why and how it is failing before making the change.
>>> I mean to say that the test is failing at run time because the correct
>>> userfaultfd syscall isn't being found with __NR_userfaultfd = 282.
>>> _NR_userfaultfd's value depends on the header. When asm-generic/unistd.h
>>> is included, its value (282) is wrong. I've tested on x86_64.
>>>
>>
>> Okay - how do you know this is wrong? can you provide more details.
>>
>> git grep _NR_userfaultfd
>> include/uapi/asm-generic/unistd.h:#define __NR_userfaultfd 282
>> include/uapi/asm-generic/unistd.h:__SYSCALL(__NR_userfaultfd,
>> sys_userfaultfd)
>> tools/include/uapi/asm-generic/unistd.h:#define __NR_userfaultfd 282
>>
>>> The fix is simple. Add the correct header which has _NR_userfaultfd =
>>> 323.
> 
> grep -rnIF "#define __NR_userfaultfd"
> tools/include/uapi/asm-generic/unistd.h:681:#define __NR_userfaultfd 282
> arch/x86/include/generated/uapi/asm/unistd_32.h:374:#define
> __NR_userfaultfd 374
> arch/x86/include/generated/uapi/asm/unistd_64.h:327:#define
> __NR_userfaultfd 323
> arch/x86/include/generated/uapi/asm/unistd_x32.h:282:#define
> __NR_userfaultfd (__X32_SYSCALL_BIT + 323)
> arch/arm/include/generated/uapi/asm/unistd-eabi.h:347:#define
> __NR_userfaultfd (__NR_SYSCALL_BASE + 388)
> arch/arm/include/generated/uapi/asm/unistd-oabi.h:359:#define
> __NR_userfaultfd (__NR_SYSCALL_BASE + 388)
> include/uapi/asm-generic/unistd.h:681:#define __NR_userfaultfd 282
> 
> The number is dependent on the architecture. The above data shows that:
> x86	374
> x86_64	323

Correct and the generated header files do the right thing and it is good to
include them as this patch does.

This is a good find and fix. I wish you explained this in your changelog.
Please add more details when you send v2.

There could be other issues lurking based on what I found.

The other two files are the problem where they hard code it to 282 without
taking the __NR_SYSCALL_BASE for the arch into consideration:

tools/include/uapi/asm-generic/unistd.h:681:#define __NR_userfaultfd 282
include/uapi/asm-generic/unistd.h:681:#define __NR_userfaultfd 282

> 
> I'm unable to find the history of why it is set to 282 in unistd.h and
> when this problem happened.

According to git history it is added in the following commit to
include/uapi/asm-generic/unistd.h:

09f7298100ea9767324298ab0c7979f6d7463183
Subject: [PATCH] userfaultfd: register uapi generic syscall (aarch64)

and it is added in the following commit to tools/include/uapi/asm-generic/unistd.h
34b009cfde2b8ce20a69c7bfd6bad4ce0e7cd970
Subject: [PATCH] tools include: Grab copies of arm64 dependent unistd.h files

I think, the above defines from include/uapi/asm-generic/unistd.h and
tools/include/uapi/asm-generic/unistd.h should be removed.

Maybe others familiar with userfaultfd can determine the best course of action.
We might have other NR_ defines in these two files that are causing problems
for tests and tools that we haven't uncovered yet.

thanks,
-- Shuah