Re: [PATCH 2/4] staging/lustre/obdclass: read jobid from proc

[Peng Tao]

On Wed, Oct 30, 2013 at 10:20 PM, Greg Kroah-Hartman
<gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> On Wed, Oct 30, 2013 at 09:31:24PM +0800, Peng Tao wrote:
>> On Wed, Oct 30, 2013 at 9:21 PM, Greg Kroah-Hartman
>> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>> > On Wed, Oct 30, 2013 at 07:30:34PM +0800, Peng Tao wrote:
>> >> so that we can get rid of cfs_get_environ() that needs access_process_vm() that
>> >> is a core mm function and is not available on some architectures.
>> >>
>> >> Reviewed-by: Niu Yawei <yawei.niu@xxxxxxxxx>
>> >> Signed-off-by: Peng Tao <bergwolf@xxxxxxxxx>
>> >> Signed-off-by: Andreas Dilger <andreas.dilger@xxxxxxxxx>
>> >> ---
>> >>  drivers/staging/lustre/lustre/obdclass/class_obd.c |  119 +++++++++++++++++++-
>> >>  1 file changed, 115 insertions(+), 4 deletions(-)
>> >>
>> >> diff --git a/drivers/staging/lustre/lustre/obdclass/class_obd.c b/drivers/staging/lustre/lustre/obdclass/class_obd.c
>> >> index b1024a6..99ce863 100644
>> >> --- a/drivers/staging/lustre/lustre/obdclass/class_obd.c
>> >> +++ b/drivers/staging/lustre/lustre/obdclass/class_obd.c
>> >> @@ -96,8 +96,117 @@ EXPORT_SYMBOL(obd_dirty_transit_pages);
>> >>  char obd_jobid_var[JOBSTATS_JOBID_VAR_MAX_LEN + 1] = JOBSTATS_DISABLE;
>> >>  EXPORT_SYMBOL(obd_jobid_var);
>> >>
>> >> -/* Get jobid of current process by reading the environment variable
>> >> - * stored in between the "env_start" & "env_end" of task struct.
>> >> +static char *self_environ_file = "/proc/self/environ";
>> >
>> > Heh, no, that's not ok at all.
>> >
>> > This is a _huge_ sign that you are doing something wrong in your driver
>> > if you need something that isn't exported, or that you have to dig out
>> > of proc.
>> >
>> Lustre has a functionality that lets applications specify a global
>> jobid by setting the same environment variable across the cluster and
>> thus server can trace IO that is issued by different job. Currently it
>> is implemented by having a private version of cfs_access_process_vm
>> because access_process_vm is not exported. The patch intends to change
>> it to reading from proc. If we cannot dig it out from proc either, how
>> can we keep the functionality then? Please advise.
>
> Kernel code should not care about environment variables.  Just because
> you created a crazy requirement, does not mean it is a valid one, nor
> that I have to be the one to tell you how to implement it.
>
> I'm never going to take code that parses proc files, you all know better
> than that.
I see. I'll drop the patchset and keep the cfs_access_process_vm code
until better option is found.

Thanks,
Tao
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/