Joseph Qi
2015-Dec-08 06:15 UTC
[Ocfs2-devel] Buffer read will get starvation in case reading/writing the same file from different nodes concurrently
On 2015/12/8 12:51, Eric Ren wrote:> Hi, > > On Tue, Dec 08, 2015 at 11:55:18AM +0800, joseph wrote: >> Hi Gang, >> Eric and I have discussed this case before. >> Using NONBLOCK here is because there is a lock inversion between inode >> lock and page lock. You can refer to the comments of >> ocfs2_inode_lock_with_page for details. >> Actually I have found that NONBLOCK mode is only used in lock inversion >> cases. > Yes, that comments can helps;-) I try to explain it, if any problem please correct me. > > On node 1, when calling ocfs2_inode_lock_with_page(), thread A(likely doing reading) has > loced this page. Before thread A stepping into ocfs2_inode_lock_full(), on node 2, thread B(doing > writing) required EX lock on the same inode, if lockres->l_level is PR, so lockres needs to > been downconverted PR->NL on node1, i.e. ocfs2_blocking_ast()->ocfs2_wake_downconvert_thread(). > > On node 1, if tread ocfs2dc proceeds like this: > ocfs2_downconvert_thread() > ocfs2_downconvert_thread_do_work() > ocfs2_process_blocked_lock() > ocfs2_unblock_lock() > lockres->l_ops->downconvert_worker(lockres, blocking) > ocfs2_data_convert_worker() { > ... > 3557 if (blocking == DLM_LOCK_EX) { > 3558 truncate_inode_pages(mapping, 0); > 3559 } else { > 3560 /* We only need to wait on the I/O if we're not also > 3561 * truncating pages because truncate_inode_pages waits > 3562 * for us above. We don't truncate pages if we're > 3563 * blocking anything < EXMODE because we want to keep > 3564 * them around in that case. */ > 3565 filemap_fdatawait(mapping); > 3566 } > ... > } > We can see truncate_inode_pages() > truncate_inode_pages_range() > trylock_page()/lock_page()/find_lock_page() > > if thread A call ocfs2_inode_lock_full() in BLOCK way, there's a window that flag > OCFS2_LOCK_BLOCED and ->l_blocking=EX have been set by BAST like ocfs2_blocking_ast()-> > ocfs2_generic_handle_bast(), so thread A may be blocked. > > Now, ocfs2dc wants page lock, but thread A hold page lock and has been blocked because of > thread B. Deadlock occurs, right? Again, please correct me if any.Yes, you are right.> > So, the author had to use UNBLOCK way here thus create livelock problem. But, I cannot understand > very clearly how livelock come up. Joseph or anyone else, could you please elaborate it?Firstly it takes inode lock in NONBLOCK way, if it couldn't be fulfilled currently and returns -EAGAIN, then unlock page and make downconvert go on. This is to release the deadlock. But we do have to take page lock and inode lock to safely accomplish the request. So it has to return to VFS and retry with page lock. Here taking inode lock and then unlocking it immediately in BLOCK way wants to cache the inode and increase the chance of the next retry. But if another node successfully takes the inode lock again immediately after this node unlocks inode, the flow you described above happens again and this node still couldn't be fulfilled. And so forth.> > Thanks, > Eric > >> >> Thanks, >> Joseph >> >> On 2015/12/8 11:21, Gang He wrote: >>> Hello Guys, >>> >>> There is a issue from the customer, who is complaining that buffer reading sometimes lasts too much time ( 1 - 10 seconds) in case reading/writing the same file from different nodes concurrently. >>> According to the demo code from the customer, we also can reproduce this issue at home (run the test program under SLES11SP4 OCFS2 cluster), actually this issue can be reproduced in openSuSe 13.2 (more newer code), but in direct-io mode, this issue will disappear. >>> Base on my investigation, the root cause is the buffer-io using cluster-lock is different from direct-io, I do not know why buffer-io use cluster-lock like this way. >>> the code details are as below, >>> in aops.c file, >>> 281 static int ocfs2_readpage(struct file *file, struct page *page) >>> 282 { >>> 283 struct inode *inode = page->mapping->host; >>> 284 struct ocfs2_inode_info *oi = OCFS2_I(inode); >>> 285 loff_t start = (loff_t)page->index << PAGE_CACHE_SHIFT; >>> 286 int ret, unlock = 1; >>> 287 >>> 288 trace_ocfs2_readpage((unsigned long long)oi->ip_blkno, >>> 289 (page ? page->index : 0)); >>> 290 >>> 291 ret = ocfs2_inode_lock_with_page(inode, NULL, 0, page); <<== this line >>> 292 if (ret != 0) { >>> 293 if (ret == AOP_TRUNCATED_PAGE) >>> 294 unlock = 0; >>> 295 mlog_errno(ret); >>> 296 goto out; >>> 297 } >>> >>> in dlmglue.c file, >>> 2 int ocfs2_inode_lock_with_page(struct inode *inode, >>> 2443 struct buffer_head **ret_bh, >>> 2444 int ex, >>> 2445 struct page *page) >>> 2446 { >>> 2447 int ret; >>> 2448 >>> 2449 ret = ocfs2_inode_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK); <<== there, why using NONBLOCK mode to get the cluster lock? this way will let reading IO get starvation. >>> 2450 if (ret == -EAGAIN) { >>> 2451 unlock_page(page); >>> 2452 if (ocfs2_inode_lock(inode, ret_bh, ex) == 0) >>> 2453 ocfs2_inode_unlock(inode, ex); >>> 2454 ret = AOP_TRUNCATED_PAGE; >>> 2455 } >>> 2456 >>> 2457 return ret; >>> 2458 } >>> >>> If you know the background behind the code, please tell us, why not use block way to get the lock in reading a page, then reading IO will get the page fairly when there is a concurrent writing IO from the other node. >>> Second, I tried to modify that line from OCFS2_LOCK_NONBLOCK to 0 (switch to blocking way), the reading IO will not be blocked too much time (can erase the customer's complaining), but a new problem arises, sometimes the reading IO and writing IO get a dead lock (why dead lock? I am looking at). >>> >>> >>> Thanks >>> Gang >>> >>> >>> >>> . >>> >> >> >> >> _______________________________________________ >> Ocfs2-devel mailing list >> Ocfs2-devel at oss.oracle.com >> https://oss.oracle.com/mailman/listinfo/ocfs2-devel >> > > _______________________________________________ > Ocfs2-devel mailing list > Ocfs2-devel at oss.oracle.com > https://oss.oracle.com/mailman/listinfo/ocfs2-devel > > . >
Eric Ren
2015-Dec-08 08:26 UTC
[Ocfs2-devel] Buffer read will get starvation in case reading/writing the same file from different nodes concurrently
Hi Joseph, On Tue, Dec 08, 2015 at 02:15:17PM +0800, joseph wrote:> On 2015/12/8 12:51, Eric Ren wrote: > > Hi, > > > > On Tue, Dec 08, 2015 at 11:55:18AM +0800, joseph wrote: > >> Hi Gang, > >> Eric and I have discussed this case before. > >> Using NONBLOCK here is because there is a lock inversion between inode > >> lock and page lock. You can refer to the comments of > >> ocfs2_inode_lock_with_page for details. > >> Actually I have found that NONBLOCK mode is only used in lock inversion > >> cases. > > Yes, that comments can helps;-) I try to explain it, if any problem please correct me. > > > > On node 1, when calling ocfs2_inode_lock_with_page(), thread A(likely doing reading) has > > loced this page. Before thread A stepping into ocfs2_inode_lock_full(), on node 2, thread B(doing > > writing) required EX lock on the same inode, if lockres->l_level is PR, so lockres needs to > > been downconverted PR->NL on node1, i.e. ocfs2_blocking_ast()->ocfs2_wake_downconvert_thread(). > > > > On node 1, if tread ocfs2dc proceeds like this: > > ocfs2_downconvert_thread() > > ocfs2_downconvert_thread_do_work() > > ocfs2_process_blocked_lock() > > ocfs2_unblock_lock() > > lockres->l_ops->downconvert_worker(lockres, blocking) > > ocfs2_data_convert_worker() { > > ... > > 3557 if (blocking == DLM_LOCK_EX) { > > 3558 truncate_inode_pages(mapping, 0); > > 3559 } else { > > 3560 /* We only need to wait on the I/O if we're not also > > 3561 * truncating pages because truncate_inode_pages waits > > 3562 * for us above. We don't truncate pages if we're > > 3563 * blocking anything < EXMODE because we want to keep > > 3564 * them around in that case. */ > > 3565 filemap_fdatawait(mapping); > > 3566 } > > ... > > } > > We can see truncate_inode_pages() > > truncate_inode_pages_range() > > trylock_page()/lock_page()/find_lock_page() > > > > if thread A call ocfs2_inode_lock_full() in BLOCK way, there's a window that flag > > OCFS2_LOCK_BLOCED and ->l_blocking=EX have been set by BAST like ocfs2_blocking_ast()-> > > ocfs2_generic_handle_bast(), so thread A may be blocked. > > > > Now, ocfs2dc wants page lock, but thread A hold page lock and has been blocked because of > > thread B. Deadlock occurs, right? Again, please correct me if any. > Yes, you are right. > > > > So, the author had to use UNBLOCK way here thus create livelock problem. But, I cannot understand > > very clearly how livelock come up. Joseph or anyone else, could you please elaborate it? > Firstly it takes inode lock in NONBLOCK way, if it couldn't be fulfilled > currently and returns -EAGAIN, then unlock page and make downconvert go > on. This is to release the deadlock. > But we do have to take page lock and inode lock to safely accomplish the > request. So it has to return to VFS and retry with page lock. Here > taking inode lock and then unlocking it immediately in BLOCK way wantsThanks! But unfortunately in our case, it oftern takes long time waiting here for inode lock, for some reasons I don't know yet. This is likely the cause why ocfs2_readpage() takes long. I really want to know why to lock & unlock inode here? I'm glad you mention the reason, but is it the only one?> to cache the inode and increase the chance of the next retry. But ifAny reason to cache the inode? And how this can increase the chance of retry? It's really interesting... Please elaborate more;-) Thanks, Eric> another node successfully takes the inode lock again immediately after > this node unlocks inode, the flow you described above happens again and > this node still couldn't be fulfilled. And so forth. > > > > > Thanks, > > Eric > > > >> > >> Thanks, > >> Joseph > >> > >> On 2015/12/8 11:21, Gang He wrote: > >>> Hello Guys, > >>> > >>> There is a issue from the customer, who is complaining that buffer reading sometimes lasts too much time ( 1 - 10 seconds) in case reading/writing the same file from different nodes concurrently. > >>> According to the demo code from the customer, we also can reproduce this issue at home (run the test program under SLES11SP4 OCFS2 cluster), actually this issue can be reproduced in openSuSe 13.2 (more newer code), but in direct-io mode, this issue will disappear. > >>> Base on my investigation, the root cause is the buffer-io using cluster-lock is different from direct-io, I do not know why buffer-io use cluster-lock like this way. > >>> the code details are as below, > >>> in aops.c file, > >>> 281 static int ocfs2_readpage(struct file *file, struct page *page) > >>> 282 { > >>> 283 struct inode *inode = page->mapping->host; > >>> 284 struct ocfs2_inode_info *oi = OCFS2_I(inode); > >>> 285 loff_t start = (loff_t)page->index << PAGE_CACHE_SHIFT; > >>> 286 int ret, unlock = 1; > >>> 287 > >>> 288 trace_ocfs2_readpage((unsigned long long)oi->ip_blkno, > >>> 289 (page ? page->index : 0)); > >>> 290 > >>> 291 ret = ocfs2_inode_lock_with_page(inode, NULL, 0, page); <<== this line > >>> 292 if (ret != 0) { > >>> 293 if (ret == AOP_TRUNCATED_PAGE) > >>> 294 unlock = 0; > >>> 295 mlog_errno(ret); > >>> 296 goto out; > >>> 297 } > >>> > >>> in dlmglue.c file, > >>> 2 int ocfs2_inode_lock_with_page(struct inode *inode, > >>> 2443 struct buffer_head **ret_bh, > >>> 2444 int ex, > >>> 2445 struct page *page) > >>> 2446 { > >>> 2447 int ret; > >>> 2448 > >>> 2449 ret = ocfs2_inode_lock_full(inode, ret_bh, ex, OCFS2_LOCK_NONBLOCK); <<== there, why using NONBLOCK mode to get the cluster lock? this way will let reading IO get starvation. > >>> 2450 if (ret == -EAGAIN) { > >>> 2451 unlock_page(page); > >>> 2452 if (ocfs2_inode_lock(inode, ret_bh, ex) == 0) > >>> 2453 ocfs2_inode_unlock(inode, ex); > >>> 2454 ret = AOP_TRUNCATED_PAGE; > >>> 2455 } > >>> 2456 > >>> 2457 return ret; > >>> 2458 } > >>> > >>> If you know the background behind the code, please tell us, why not use block way to get the lock in reading a page, then reading IO will get the page fairly when there is a concurrent writing IO from the other node. > >>> Second, I tried to modify that line from OCFS2_LOCK_NONBLOCK to 0 (switch to blocking way), the reading IO will not be blocked too much time (can erase the customer's complaining), but a new problem arises, sometimes the reading IO and writing IO get a dead lock (why dead lock? I am looking at). > >>> > >>> > >>> Thanks > >>> Gang > >>> > >>> > >>> > >>> . > >>> > >> > >> > >> > >> _______________________________________________ > >> Ocfs2-devel mailing list > >> Ocfs2-devel at oss.oracle.com > >> https://oss.oracle.com/mailman/listinfo/ocfs2-devel > >> > > > > _______________________________________________ > > Ocfs2-devel mailing list > > Ocfs2-devel at oss.oracle.com > > https://oss.oracle.com/mailman/listinfo/ocfs2-devel > > > > . > > > > >