Messages in this thread | | | Date | Tue, 23 Nov 2010 09:36:31 +0800 | Subject | Re: [PATCH] arch/tile: fix rwlock so would-be write lockers don't block new readers | From | Cypher Wu <> |
| |
2010/11/22 Chris Metcalf <cmetcalf@tilera.com>: > On 11/22/2010 12:39 AM, Cypher Wu wrote: >> 2010/11/15 Chris Metcalf <cmetcalf@tilera.com>: >>> This avoids a deadlock in the IGMP code where one core gets a read >>> lock, another core starts trying to get a write lock (thus blocking >>> new readers), and then the first core tries to recursively re-acquire >>> the read lock. >>> >>> We still try to preserve some degree of balance by giving priority >>> to additional write lockers that come along while the lock is held >>> for write, so they can all complete quickly and return the lock to >>> the readers. >>> >>> Signed-off-by: Chris Metcalf <cmetcalf@tilera.com> >>> --- >>> This should apply relatively cleanly to 2.6.26.7 source code too. >>> [...] >> >> I've finished my business trip and tested that patch for more than an >> hour and it works. The test is still running now. >> >> But it seems there still has a potential problem: we used ticket lock >> for write_lock(), and if there are so many write_lock() occurred, is >> 256 ticket enough for 64 or even more cores to avoiding overflow? >> Since is we try to write_unlock() and there's already write_lock() >> waiting we'll only adding current ticket. > > This is OK, since each core can issue at most one (blocking) write_lock(), > and we have only 64 cores. Future >256 core machines will be based on > TILE-Gx anyway, which doesn't have the 256-core limit since it doesn't use > the spinlock_32.c implementation. > > -- > Chris Metcalf, Tilera Corp. > http://www.tilera.com > >
Say, if core A try to write_lock() rwlock and current_ticket_ is 0 and it write next_ticket_ to 1, when it processing the lock, core B try to write_lock() again and write next_ticket_ to 2, then when A write_unlock() it seen that (current_ticket_+1) is not equal to next_ticket_, so it increment current_ticket_, and core B get the lock. If core A try write_lock again before core B write_unlock, it will increment next_ticket_ to 3. And so on. This may rarely happened, I've tested it yesterday for several hours it goes very well under pressure.
-- Cyberman Wu http://www.meganovo.com
| |