On 18/09/2020 14.13, Petr Mladek wrote:
> On Fri 2020-09-18 08:16:37, Rasmus Villemoes wrote:
>> On 17/09/2020 15.16, John Ogness wrote:
>>
>>> if (dev->class)
>>> subsys = dev->class->name;
>>> else if (dev->bus)
>>> subsys = dev->bus->name;
>>> else
>>> - return 0;
>>> + return;
>>>
>>> - pos += snprintf(hdr + pos, hdrlen - pos, "SUBSYSTEM=%s", subsys);
>>> - if (pos >= hdrlen)
>>> - goto overflow;
>>> + snprintf(dev_info->subsystem, sizeof(dev_info->subsystem), subsys);
>>
>> It's unlikely that subsys would contain a %, but this will be yet
>> another place to spend brain cycles ignoring if doing static analysis.
>> So can we not do this. Either of strXcpy() for X=s,l will do the same
>> thing, and likely faster.
>
> Good point! Better be on the safe size in a generic printk() API.
>
> Well, I am afraid that this would be only small drop in a huge lake.
> class->name and bus->name seems to be passed to %s in so many
> *print*() calls all over the kernel code.
So what? printf("%s", some_string_that_might_contain_percent_chars) is
not a problem.
printf(some_string_that_might_contain_percent_chars) is.
And yes, one could do
snprintf(dev_info->subsystem, sizeof(dev_info->subsystem), "%s", subsys);
but one might as well avoid the snprintf overhead and use one of the
strX functions that have the exact same semantics (copy as much as
there's room for, ensure nul-termination).
Rasmus
_______________________________________________
kexec mailing list
kexec@xxxxxxxxxxxxxxxxxxx
http://lists.infradead.org/mailman/listinfo/kexec
