linux-kernel - Re: [PATCH v1 1/2] perf record: Add 8-byte aligned event type PERF_RECORD

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: <CAP-5=fU4rVkTS07Uq8g9roO4kXq_z2R0CgMX55YFXFWMHzWTGw@mail.gmail.com>
Date: Wed, 26 Feb 2025 22:20:36 -0800
From: Ian Rogers <irogers@...gle.com>
To: Andi Kleen <ak@...ux.intel.com>
Cc: Chun-Tse Shao <ctshao@...gle.com>, linux-kernel@...r.kernel.org, peterz@...radead.org, 
	mingo@...hat.com, acme@...nel.org, namhyung@...nel.org, mark.rutland@....com, 
	alexander.shishkin@...ux.intel.com, jolsa@...nel.org, adrian.hunter@...el.com, 
	kan.liang@...ux.intel.com, terrelln@...com, leo.yan@....com, 
	dvyukov@...gle.com, james.clark@...aro.org, christophe.leroy@...roup.eu, 
	ben.gainey@....com, linux-perf-users@...r.kernel.org
Subject: Re: [PATCH v1 1/2] perf record: Add 8-byte aligned event type PERF_RECORD_COMPRESSED2

On Wed, Feb 26, 2025 at 10:04 PM Andi Kleen <ak@...ux.intel.com> wrote:
>
> On Wed, Feb 26, 2025 at 09:34:06PM -0800, Chun-Tse Shao wrote:
> > The original PERF_RECORD_COMPRESS is not 8-byte aligned, which can cause
> > asan runtime error:
>
> It seems pointless. Most architectures have cheap unaligned accesses
> these days.
>
> Just disable that error?

The perf_event_header in perf_event.h is:
```
struct perf_event_header {
__u32 type;
__u16 misc;
__u16 size;
};
```
so it is assuming at least 4-byte alignment. 8-byte alignment is
assumed in many places in tools/lib/perf/include/perf/event.h. We pad
events to ensure the alignment in about 30 places already:
```
$ grep -r PERF_ALIGN tools/perf|grep u64|wc -l
32
```
Having sanitizers I think is a must, if we allow unaligned events we'd
need to introduce helper functions or memcpys to workaround the
unaligned undefined behavior. I think the padding is a less worse
alternative and one that was already picked.

Thanks,
Ian