Re: [RFC][PATCH 4/6][v3] Define siginfo_from_ancestor_ns()

[Sukadev Bhattiprolu <sukadev@linux.vnet.ibm.com>]

Oleg Nesterov [oleg@redhat.com] wrote:
| On 12/20, Sukadev Bhattiprolu wrote:
| >
| > + * TODO:
| > + * 	  Making SI_ASYNCIO a kernel signal could make this less hacky.
| > + */
| > +#ifdef CONFIG_PID_NS
| > +static inline int siginfo_from_user(siginfo_t *info)
| > +{
| > +	if (!is_si_special(info) && SI_FROMUSER(info) &&
| 
| OK, if we can trust SI_FROMUSER(), then it is better, i agree.
| 
| I was worried about in-kernel usage of .si_code <= 0 ...
| 
| > +				info->si_code != SI_ASYNCIO)
| 
| but this is horrible, imho.

I am beginning to accept that some amount of ugliness is inevitable
here :-) I tried to dig through history of SI_ASYNCIO, but did not
find any changes to its definition in siginfo.h in 6 years.

| 
| OK, if we can't change the ABI, then perhaps we can change
| kill_pid_info_as_uid() to not send the fatal signals to UNKILLABLE
| task? This helper is strange and ugly anyway,
| 
| 
| To clarify, I do not blame the patch itself, and I do not suggest
| to do this right now.

By 'to do this' I assume you are referring to the kill_pid_info_as_uid()
change above ?

IOW, ugly as it is, can we go with the siginfo_from_user() as in the patch ?