Re: [RFC PATCH v2 2/3] security: Add PR_CAPBSET_DROP_MASK for process-wide bounding-set drops

From: Jinjie Ruan

Date: Tue Nov 17 2026 - 02:09:14 EST




在 2026/11/10 2:46, Serge E. Hallyn 写道:
> On Mon, Nov 09, 2026 at 07:00:51PM +0100, Oleg Nesterov wrote:
>> On 11/09, Serge E. Hallyn wrote:
>>>
>>> Hi Oleg,
>>>
>>> please see below. Do you have any issues with this use of sighand?
>>
>> Hi Serge,
>>
>> I already tried to look at this patch, but sorry, can't really comment.
>>
>> I'll try to take another look, but can't promise much, I don't understand
>> the problem space (I could never reallly understand capabilites in details).
>
> Thanks. I'm mainly wondering about the storing of this pending capability
> bounding mask in signal_struct, whether that might have any unexpected

Hi Serge,

rlimit is also placed inside the signal struct to implement
process-level restrictions.

I think the different lifetimes of cred and signal struct might be an
issue, but since drop only decreases and never increases, the result
seen by the user will only be stricter.

Link: https://lwn.net/Articles/103658/
Link:
https://github.com/torvalds/linux/commit/31180071ee5e6cc6ff4d036d655c556f582f74e4

Best regards,
Jinjie

> side effects.