[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [PATCH 8/8] storage: Guess the parent if it's not specified for vHBA

On 02/06/2013 11:32 AM, John Ferlan wrote:

>>>> +
>>>> +        ignore_value(sscanf(entry->d_name, "host%d",&host));
>>> Why ignore_value()?  If == -1, then host is undefined - could be
>>> something that results in the following being successful..
>> I don't think it's possible to return -1, as all the entries
>> under SYSFS_FC_HOST_PATH should be "hostN". Entry "." and ".."
>> are already skipped.
> sscanf can return -1 for other errors (ENOMEM included) - I would think
> it's safer to check and fail than assume anything.

sscanf should NEVER be used to parse raw "%d".  POSIX says that the
result is undefined on integer overflow (sscanf is not required to fail
in that case, yet you are not guaranteed if you got MAX_INT, wraparound
modulo 2**32, or some other weird behavior).  Something like
sscanf("%5d") for parsing at most five digits is slightly more tolerable
because it prevents you from getting to the overflow situation, although
I still think that using sscanf to parse integers is dangerous.  And
even in the cases where using sscanf is safe (fixed-length parsing not
subject to integer overflow), you should NEVER ignore failure, and you
should always end with a %n to ensure that you parsed as much as you
were expecting to parse.  sscanf() is so difficult to correctly use
directly that I recommend that we avoid adding any new uses of it into
libvirt (I'd like to turn on the 'make syntax-check' rule that forbids
*scanf entirely, but we'd have to first convert our existing uses into
alternative code).

I'd much rather see virStrToLong_i() used here.

Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

Attachment: signature.asc
Description: OpenPGP digital signature

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]