lkml.org 
[lkml]   [2017]   [Apr]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH] FS-Cache: print hexadecimal value for special cookies type
On Tue, 25 Apr 2017 00:47:41 +0300
Andy Shevchenko <andy.shevchenko@gmail.com> wrote:

> On Sat, Apr 22, 2017 at 2:09 AM, Jérémy Lefaure
> <jeremy.lefaure@lse.epita.fr> wrote:
> > On Fri, 21 Apr 2017 22:41:54 +0300
> > Andy Shevchenko <andy.shevchenko@gmail.com> wrote:
> >> On Wed, Apr 19, 2017 at 11:38 PM, Jérémy Lefaure
> >> <jeremy.lefaure@lse.epita.fr> wrote:
> >> > When building object-list.o, gcc 6 raises a warning on the sprintf call
> >> > in fscache_objlist_show:
> >> >
> >> > CC fs/fscache/object-list.o
> >> > fs/fscache/object-list.c: In function ‘fscache_objlist_show’:
> >> > fs/fscache/object-list.c:265:19: warning: ‘sprintf’ may write a
> >> > terminating nul past the end of the destination [-Wformat-overflow=]
> >> > sprintf(_type, "%02u", cookie->def->type);
> >> > ^~~~~~
> >> > fs/fscache/object-list.c:265:4: note: ‘sprintf’ output between 3 and 4
> >> > bytes into a destination of size 3
> >> > sprintf(_type, "%02u", cookie->def->type);
> >> > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> >> >
> >> > Moreover, the documentation says that we should have an hex value for
> >> > special cookies (see Documentation/filesystems/caching/fscache.txt).
> >> >
> >> > Printing hexadecimal value for special cookies fixes the overflow
> >> > warning and complies with the documentation.
> >>
> >> If this is used by some user space tool the safest fix is to print in
> >> BCD and fix documentation.
> >>
> > Are you talking about the Binary Coded Decimal format ?
>
> Yes.
>
> > I don't get why
> > it would be better. If we change the printed format, it would break
> > user space tool anyway, right ?
>
> If userspace tool is bound to the format it might go crazy about hex
> digits in the cookie.
>
> Is that cookie important somehow? (for example, user space might
> request something based on it?) It might be a case when you got let
> say '75' and then ask for 117 (0x75), however real value is 75.
>
Ok I get it. But changing the format can break the userland tool
anyway, can't it ? For example if a userland tool expects the value 75,
it won't read it properpy in BCD.

> So, if it's okay in this, perhaps paragraph explaining side effects in
> commit message would be good to have/
>
Maybe the best fix would be to increase the size of the _type buffer to
4 so the unsigned byte can fit in it. And change the documentation to
decimal value. What do you think about this solution ?

Thanks,

\
 
 \ /
  Last update: 2017-04-25 23:37    [W:0.064 / U:0.004 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site