Re: [PATCH] sysfs: clamp show() return value in sysfs_kf_read()

[David Laight]

On Thu, 21 May 2026 08:18:32 +0200
Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:

> On Wed, May 20, 2026 at 08:19:34AM -1000, Tejun Heo wrote:
> > Hello,
> > 
> > Two nits:
> > 
> > - Buffer is atomic_write_len ?: PAGE_SIZE, so probably better to clamp
> >   to that than hardcode PAGE_SIZE.  
> 
> Where is that check at?  And sysfs_kf_seq_show() doesn't check it this
> way, should that change?

It would be better to always fix the 'show' buffer at PAGE_SIZE (or 4k).

The only place that uses is different value is the cgroup code and I
think it needs to support longer writes (100 + 6 * NR_CPUS).
I've not looked at how it handles the matching reads - I presume they exist.
If the data is binary, or it uses seq_printf() then it can just support
a longer dynamically allocated buffer.

-- David

> 
> > - pr_warn() instead of bare printk()?  
> 
> This is copying the message in sysfs_kf_seq_show() that has the same
> exact format.  I will send a follow-on patch to make both the same.
> 
> thanks,
> 
> greg k-h
>