lkml.org 
[lkml]   [2013]   [Mar]   [14]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH 11/12] rwsem: wake all readers when first waiter is a reader
From
On Thu, Mar 14, 2013 at 4:39 AM, Peter Hurley <peter@hurleysoftware.com> wrote:
> On Thu, 2013-03-14 at 00:03 -0700, Michel Lespinasse wrote:
>> > CPU 0 | CPU 1
>> > |
>> > | down_write()
>> >
>> > ... CPU 1 has the write lock for the semaphore.
>> > Meanwhile, 1 or more down_read(s) are attempted and fail;
>> > these are put on the wait list.
>>
>> I'm not sure of the relevance of these other down_read() calls -
>> please note that as these extra readers are put on the wait list,
>> their +read_bias adjustments are canceled so that the count value
>> ends up at write_bias + waiting_bias (for a total active count of 1)
>
> The relevance of the waiting readers is that when CPU 1 drops the writer
> ___and grabs the spin lock___, it then wakes up these already-waiting
> readers (CPU 0 is still parked in down_read_failed() waiting to acquire
> the spin lock).
>
> When CPU 1 wakes up these readers, the sem count goes > 0 and the
> waiting list is emptied. CPU 1 then drops the spin lock and leaves
> up_write().

Ah, correct. So the race you noticed is real, and the one I found is
just a more complicated way to hit it.

I'll add a fix in v2 of this patch series (probably later today).

BTW, you seem to have been paying pretty close attention, would you
mind giving appropriate Reviewed-by tags when I send v2 ?


\
 
 \ /
  Last update: 2013-03-14 17:01    [W:0.090 / U:0.860 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site