Re: [PATCH 1/2] staging: lustre: lprocfs: Use kstrtouint_from_user

From: Dilger, Andreas
Date: Thu May 18 2017 - 23:03:08 EST


On May 18, 2017, at 17:13, Mathias Rav <mathiasrav@xxxxxxxxx> wrote:
>
> On Thu, 18 May 2017 14:48:25 +0000
> "Dilger, Andreas" <andreas.dilger@xxxxxxxxx> wrote:
>
>> On May 18, 2017, at 15:53, Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>>>
>>> On Thu, May 04, 2017 at 12:13:38PM -0400, Mathias Rav wrote:
>>>> Prefer kstrtouint_from_user to copy_from_user+simple_strtoul.
>>>>
>>>> The helper function lprocfs_wr_uint() is only used to implement
>>>> "dump_granted_max" in debugfs.
>>>>
>>>> Note the slight change in semantics: The previous implementation using
>>>> simple_strtoul allows garbage after the number, whereas kstrtox only allows
>>>> a trailing line break. The previous implementation allowed a write of zero
>>>> bytes whereas kstrtox will return -EINVAL. Since this only affects a single
>>>> debugfs endpoint, this should be a permissible slight change of semantics
>>>> in exchange for 18 fewer lines of code.
>>>>
>>>> Signed-off-by: Mathias Rav <mathiasrav@xxxxxxxxx>
>>>> ---
>>>> .../lustre/lustre/obdclass/lprocfs_status.c | 22 +---------------------
>>>> 1 file changed, 1 insertion(+), 21 deletions(-)
>>>>
>>>> diff --git a/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c b/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c
>>>> index 1ec6e3767d81..338ce34d6514 100644
>>>> --- a/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c
>>>> +++ b/drivers/staging/lustre/lustre/obdclass/lprocfs_status.c
>>>> @@ -399,27 +399,7 @@ EXPORT_SYMBOL(lprocfs_rd_uint);
>>>> int lprocfs_wr_uint(struct file *file, const char __user *buffer,
>>>> unsigned long count, void *data)
>>>> {
>>>> - unsigned *p = data;
>>>> - char dummy[MAX_STRING_SIZE + 1], *end;
>>>> - unsigned long tmp;
>>>> -
>>>> - if (count >= sizeof(dummy))
>>>> - return -EINVAL;
>>>> -
>>>> - if (count == 0)
>>>> - return 0;
>>>> -
>>>> - if (copy_from_user(dummy, buffer, count))
>>>> - return -EFAULT;
>>>> -
>>>> - dummy[count] = '\0';
>>>> -
>>>> - tmp = simple_strtoul(dummy, &end, 0);
>>>> - if (dummy == end)
>>>> - return -EINVAL;
>>>> -
>>>> - *p = (unsigned int)tmp;
>>>> - return count;
>>>> + return kstrtouint_from_user(buffer, count, 0, (unsigned int *)data);
>>>
>>> Why not just delete this whole function and have the callers make this
>>> call instead? No need for unneeded wrapper functions of core kernel
>>> calls.
>>
>> Even better, it looks like this function has no callers on the client and could just
>> be deleted entirely.
>
> No, the functions lprocfs_{rd,wr}_uint are used once through a macro:
> ldlm/ldlm_resource.c calls LPROC_SEQ_FOPS_RW_TYPE(ldlm_rw, uint)
> to define ldlm_rw_uint_seq_{show,write}, which then calls
> lprocfs_{rd,wr}_uint which in turn call seq_printf/kstrtouint_from_user.
>
> It seems like much indirection for almost no gain besides hiding
> access to ((seq_file *) file->private_data)->private in a macro.
>
> If you agree, I will prepare a patch that switches ldlm_resource to
> using the LPROC_SEQ_FOPS macro directly, which allows us to remove two
> trivial wrappers.

Please do.

Cheers, Andreas
--
Andreas Dilger
Lustre Principal Architect
Intel Corporation