linux-kernel - Re: [PATCH] ACPI / PPTT: get PPTT table in the first beginning

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: <1c9112c0-8a3e-e82e-8c88-377e3d4e9e18@huawei.com>
Date:   Wed, 21 Jul 2021 09:25:23 +0800
From:   Xiongfeng Wang <wangxiongfeng2@...wei.com>
To:     Sudeep Holla <sudeep.holla@....com>
CC:     <catalin.marinas@....com>, <will@...nel.org>, <rjw@...ysocki.net>,
        <viresh.kumar@...aro.org>, <linux-arm-kernel@...ts.infradead.org>,
        <linux-kernel@...r.kernel.org>, <linux-acpi@...r.kernel.org>,
        <guohanjun@...wei.com>
Subject: Re: [PATCH] ACPI / PPTT: get PPTT table in the first beginning

Hi Sudeep,

On 2021/7/20 21:37, Sudeep Holla wrote:
> On Tue, Jul 20, 2021 at 07:26:35PM +0800, Xiongfeng Wang wrote:
>> When I added might_sleep() in down_timeout(), I got the following
> 
> Sorry it is not clear if you are able to reproduce this issue without
> any other modifications in the mainline kernel ?

Without any modifications, the mainline kernel does not print the Calltrace. But
the risk of sleeping in atomic context still exists.

> 
>> Calltrace:
>>
>> [    8.775671] BUG: sleeping function called from invalid context at kernel/locking/semaphore.c:160
>> [    8.777070] in_atomic(): 1, irqs_disabled(): 128, non_block: 0, pid: 14, name: cpuhp/0
> 
>>>From this I guess you are adding sleep after raw_spin_lock_irqsave
> in down_timeout(kernel/locking/semaphore.c).

I add 'might_sleep()' which is used to check whether there are problems if I
sleep here. It's not a real sleep.

The Document for might_sleep is as below.
/**
 * might_sleep - annotation for functions that can sleep

 *
 * this macro will print a stack trace if it is executed in an atomic
 * context (spinlock, irq-handler, ...). Additional sections where blocking is
 * not allowed can be annotated with non_block_start() and non_block_end()
 * pairs.
 *
 * This is a useful debugging help to be able to catch problems early and not
 * be bitten later when the calling function happens to sleep when it is not
 * supposed to.
 */

> 
>>
>> It is because generic_exec_single() will disable local irq before
>> calling _init_cache_level(). _init_cache_level() use acpi_get_table() to
>> get the PPTT table, but this function could schedule out.
>>
>> To fix this issue, we use a static pointer to record the mapped PPTT
>> table in the first beginning. Later, we use that pointer to reference
>> the PPTT table in acpi_find_last_cache_level(). We also modify other
>> functions in pptt.c to use the pointer to reference PPTT table.

The main problem is that we called local_irq_save() in generic_exec_single(),
and then we called down_timeout() in acpi_os_wait_semaphore(). down_timeout()
could enter into sleep if failed to acquire the semaphore. There are risks of
sleeping in irq disabled context.

Thanks,
Xiongfeng

>>
> 
> I don't follow this change at all.
>