Re: [PATCH v3] libfs: fix error cast of negative value in simple_attr_write()

[Yicong Yang]

On 2021/6/4 0:09, Luck, Tony wrote:
> On Sat, Nov 14, 2020 at 04:09:16PM +0800, Yicong Yang wrote:
>> The attr->set() receive a value of u64, but simple_strtoll() is used
>> for doing the conversion. It will lead to the error cast if user inputs
>> a negative value.
>>
>> Use kstrtoull() instead of simple_strtoll() to convert a string got
>> from the user to an unsigned value. The former will return '-EINVAL' if
>> it gets a negetive value, but the latter can't handle the situation
>> correctly. Make 'val' unsigned long long as what kstrtoull() takes, this
>> will eliminate the compile warning on no 64-bit architectures.
>>
>> Fixes: f7b88631a897 ("fs/libfs.c: fix simple_attr_write() on 32bit machines")
>> Signed-off-by: Yicong Yang <yangyicong@xxxxxxxxxxxxx>
>> ---
>> Change since v1:
>> - address the compile warning for non-64 bit platform.
>> Change since v2:
>> Link: https://lore.kernel.org/linux-fsdevel/1605000324-7428-1-git-send-email-yangyicong@xxxxxxxxxxxxx/
>> - make 'val' unsigned long long and mentioned in the commit
>> Link: https://lore.kernel.org/linux-fsdevel/1605261369-551-1-git-send-email-yangyicong@xxxxxxxxxxxxx/
> 
> Belated error report on this. Some validation team just moved to
> v5.10 and found their error injection scripts no longer work.
> 
> They have been using:
> 
> # echo $((-1 << 12)) > /sys/kernel/debug/apei/einj/param2
> 
> to write the mask value 0xfffffffffffff000 for many years ... but
> now writing a negative value (-4096) to this file gives an EINVAL error.
> 

ok. i didn't know this usage. I was using debugfs for my driver but
found I cannot figure out whether user has entered a negative value.

> Maybe they've been taking advantage of a bug all this time? The
> comment for debugfs_create_x64() says it is for reading/writing
> an unsigned value. But when a bug fix breaks user code ... then
> we are supposed to ask whether that bug is actually a feature.
> 

ok. sounds reasonable.

> If there was a debugfs_create_s64() I might just fix einj.c to
> use that ... but there isn't :-(

yes, a debugfs_create_s64() seems better for this case.
But it's okay for me to revert this fix if we regard this bug as
a feature.

Thanks
Yicong