[REGRESSION] [BISECTED] linux-6.6.y and linux-6.12.y: proc: use the same treatment to check proc_lseek as ones for proc_read_iter et.al

[REGRESSION] [BISECTED] linux-6.6.y and linux-6.12.y: proc: use the same treatment to check proc_lseek as ones for proc_read_iter et.al

[Lars Wendler]

Hi,

the stable LTS linux kernels 6.6.102 and 6.12.42 have a regression
regarding network interface monitoring with xosview and gkrellm. Both
programs no longer show any network traffic with gkrellm even
considering all network interfaces as being in down state. I haven't
checked other LTS kernels so I cannot tell if there are more affected
kernel branches.

I have bisected the issue to the commits
33c778ea0bd0fa62ff590497e72562ff90f82b13 in 6.6.102 and
fc1072d934f687e1221d685cf1a49a5068318f34 in 6.12.42 which are both the
same change code-wise (upstream commit
ff7ec8dc1b646296f8d94c39339e8d3833d16c05).

Reverting these commits makes xosview and gkrellm "work" again as in
they both show network traffic again.

Kind regards
Lars Wendler

Re: [REGRESSION] [BISECTED] linux-6.6.y and linux-6.12.y: proc: use the same treatment to check proc_lseek as ones for proc_read_iter et.al

[Greg KH]

On Fri, Aug 15, 2025 at 07:56:16PM +0200, Lars Wendler wrote:
> Hi,
> 
> the stable LTS linux kernels 6.6.102 and 6.12.42 have a regression
> regarding network interface monitoring with xosview and gkrellm. Both
> programs no longer show any network traffic with gkrellm even
> considering all network interfaces as being in down state. I haven't
> checked other LTS kernels so I cannot tell if there are more affected
> kernel branches.
> 
> I have bisected the issue to the commits
> 33c778ea0bd0fa62ff590497e72562ff90f82b13 in 6.6.102 and
> fc1072d934f687e1221d685cf1a49a5068318f34 in 6.12.42 which are both the
> same change code-wise (upstream commit
> ff7ec8dc1b646296f8d94c39339e8d3833d16c05).
> 
> Reverting these commits makes xosview and gkrellm "work" again as in
> they both show network traffic again.

Is this also an issue in 6.16.1 and 6.17-rc1?

thanks,

greg k-h

Re: [REGRESSION] [BISECTED] linux-6.6.y and linux-6.12.y: proc: use the same treatment to check proc_lseek as ones for proc_read_iter et.al

[wangzijie]

> Hi,
> 
> the stable LTS linux kernels 6.6.102 and 6.12.42 have a regression
> regarding network interface monitoring with xosview and gkrellm. Both
> programs no longer show any network traffic with gkrellm even
> considering all network interfaces as being in down state. I haven't
> checked other LTS kernels so I cannot tell if there are more affected
> kernel branches.
> 
> I have bisected the issue to the commits
> 33c778ea0bd0fa62ff590497e72562ff90f82b13 in 6.6.102 and
> fc1072d934f687e1221d685cf1a49a5068318f34 in 6.12.42 which are both the
> same change code-wise (upstream commit
> ff7ec8dc1b646296f8d94c39339e8d3833d16c05).
> 
> Reverting these commits makes xosview and gkrellm "work" again as in
> they both show network traffic again.
> 
> Kind regards
> Lars Wendler

Hi,
Sorry for the delay, I have a rest on weekends. I think I get what happened.
We should call pde_set_flags when create proc_dir_entry, and there has omission
in net related proc file. I will fix it ASAP.