linux-kernel - Re: [PATCH] tracing: zero the pipe cpumask on alloc to avoid spurious -EBUSY

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 for Android: free password hash cracker in your pocket

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

Message-ID: <26f8fb43-6ea8-edc2-381d-3600fce261af@huawei.com>
Date:   Thu, 31 Aug 2023 21:51:18 +0800
From:   Zheng Yejian <zhengyejian1@...wei.com>
To:     Brian Foster <bfoster@...hat.com>,
        <linux-trace-kernel@...r.kernel.org>
CC:     <linux-kernel@...r.kernel.org>, <rostedt@...dmis.org>
Subject: Re: [PATCH] tracing: zero the pipe cpumask on alloc to avoid spurious
 -EBUSY

On 2023/8/31 20:55, Brian Foster wrote:
> The pipe cpumask used to serialize opens between the main and percpu
> trace pipes is not zeroed or initialized. This can result in
> spurious -EBUSY returns if underlying memory is not fully zeroed.
> This has been observed by immediate failure to read the main
> trace_pipe file on an otherwise newly booted and idle system:
> 
>   # cat /sys/kernel/debug/tracing/trace_pipe
>   cat: /sys/kernel/debug/tracing/trace_pipe: Device or resource busy
> 
> Zero the allocation of pipe_cpumask to avoid the problem.
> 
> Fixes: c2489bb7e6be ("tracing: Introduce pipe_cpumask to avoid race on trace_pipes")
> Signed-off-by: Brian Foster <bfoster@...hat.com>
> ---
> 
> Hi,
> 
> I ran into this problem just recently on one of my test VMs immediately
> after updating to a v6.5 base. A revert of the aforementioned commit
> addressed the problem. I'm not terribly familiar with the tracing code,
> but on further inspection I noticed the cpumask doesn't appear to be
> initialized anywhere. I suppose this could alternatively do a
> cpumask_clear() or whatever after allocation, but either way this
> addresses the problem for me.

Yes, pipe_cpumask must be initialized.

--

Thanks,
Zheng Yejian

> 
> Please CC on replies as I'm not subscribed to the list. Thanks.
> 
> Brian
> 
>   kernel/trace/trace.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
> index 8e64aaad5361..2656ca3b9b39 100644
> --- a/kernel/trace/trace.c
> +++ b/kernel/trace/trace.c
> @@ -9486,7 +9486,7 @@ static struct trace_array *trace_array_create(const char *name)
>   	if (!alloc_cpumask_var(&tr->tracing_cpumask, GFP_KERNEL))
>   		goto out_free_tr;
>   
> -	if (!alloc_cpumask_var(&tr->pipe_cpumask, GFP_KERNEL))
> +	if (!zalloc_cpumask_var(&tr->pipe_cpumask, GFP_KERNEL))
>   		goto out_free_tr;
>   
>   	tr->trace_flags = global_trace.trace_flags & ~ZEROED_TRACE_FLAGS;
> @@ -10431,7 +10431,7 @@ __init static int tracer_alloc_buffers(void)
>   	if (trace_create_savedcmd() < 0)
>   		goto out_free_temp_buffer;
>   
> -	if (!alloc_cpumask_var(&global_trace.pipe_cpumask, GFP_KERNEL))
> +	if (!zalloc_cpumask_var(&global_trace.pipe_cpumask, GFP_KERNEL))
>   		goto out_free_savedcmd;
>   
>   	/* TODO: make the number of buffers hot pluggable with CPUS */