[CRIU] [PATCH 2/3] cg: allow properties that aren't integers

[Tycho Andersen]

On Fri, Aug 22, 2014 at 05:00:05PM +0400, Pavel Emelyanov wrote:
> On 08/22/2014 04:49 PM, Tycho Andersen wrote:
> > On Fri, Aug 22, 2014 at 01:10:21PM +0400, Pavel Emelyanov wrote:
> >> On 08/22/2014 01:00 AM, Tycho Andersen wrote:
> >>> In particular, cpuset.cpus and cpuset.mems can both be "lists" (strings), as
> >>> well as hex integers. We don't use the result of this parse, so it is fine to delete it.
> >>>
> >>> Signed-off-by: Tycho Andersen <tycho.andersen at canonical.com>
> >>> ---
> >>>  cgroup.c | 14 +++-----------
> >>>  1 file changed, 3 insertions(+), 11 deletions(-)
> >>>
> >>> diff --git a/cgroup.c b/cgroup.c
> >>> index c5ae154..b821762 100644
> >>> --- a/cgroup.c
> >>> +++ b/cgroup.c
> >>> @@ -295,12 +295,12 @@ static inline char *strip(char *str)
> >>>  }
> >>>  
> >>>  /*
> >>> - * Currently this function only supports properties that have 1 value, under 100
> >>> - * chars
> >>> + * Currently this function only supports properties that have a string value
> >>> + * under 1024 chars.
> >>>   */
> >>>  static int read_cgroup_prop(struct cgroup_prop *property, const char *fullpath)
> >>>  {
> >>> -	char buf[100];
> >>> +	char buf[1024];
> >>>  	FILE *f;
> >>>  	char *endptr;
> >>>  
> >>> @@ -325,14 +325,6 @@ static int read_cgroup_prop(struct cgroup_prop *property, const char *fullpath)
> >>>  		return -1;
> >>>  	}
> >>>  
> >>> -	if (strtoll(buf, &endptr, 10) == LLONG_MAX)
> >>> -		strcpy(buf, "-1");
> >>> -
> >>> -	if (strcmp(endptr, "\n")) {
> >>> -		pr_perror("Failed parsing %s, with strtoll\n", buf);
> >>> -		return -1;
> >>> -	}
> >>
> >> This code is required to properly handle the memory controller's unlimited value.
> >> See commit a152c843b8af67bf8cbb3254a4b50f8e7a8b1104
> > 
> > Ah, whoops. Do we have any idea why the write fails?
> 
> AFAIU it's the kernel. It doesn't accept bug values, asking for "-1" to
> denote "unlimited", but shows the LLONG_MAX on read.

Ah, ok.

> > I guess we don't need to delete the strtoll test, just the strcmp is
> > what breaks things.
> 
> OK, will you re-send the fixed patch?

Yes, I re-sent patch 3 too, since there was a typo when I split it
apart that would have made the old patch 3 cause things to not
compile. Sorry for the confusion.

Tycho

> > Tycho
> > 
> >>> -
> >>>  	property->value = xstrdup(strip(buf));
> >>>  	if (!property->value)
> >>>  		return -1;
> >>>
> >>
> > .
> > 
>