Re: [PATCH 00/13] netfs, cifs: Fixes to retry-related code

[David Howells <dhowells@redhat.com>]

Hi Max,

Depending on what you're doing on ceph, you might need the attached patch as
well.  I managed to reproduce it by doing a git clone and kernel build on a
ceph mount with cachefiles active.

David
----
commit 6c24e7124642846b628274c845cd6de252f1dfbf
Author: David Howells <dhowells@redhat.com>
Date:   Thu Jul 10 15:02:57 2025 +0100

    netfs: Fix race between cache write completion and ALL_QUEUED being set
    
    When netfslib is issuing subrequests, the subrequests start processing
    immediately and may complete before we reach the end of the issuing
    function.  At the end of the issuing function we set NETFS_RREQ_ALL_QUEUED
    to indicate to the collector that we aren't going to issue any more subreqs
    and that it can do the final notifications and cleanup.
    
    Now, this isn't a problem if the request is synchronous
    (NETFS_RREQ_OFFLOAD_COLLECTION is unset) as the result collection will be
    done in-thread and we're guaranteed an opportunity to run the collector.
    
    However, if the request is asynchronous, collection is primarily triggered
    by the termination of subrequests queuing it on a workqueue.  Now, a race
    can occur here if the app thread sets ALL_QUEUED after the last subrequest
    terminates.
    
    This can happen most easily with the copy2cache code (as used by Ceph)
    where, in the collection routine of a read request, an asynchronous write
    request is spawned to copy data to the cache.  Folios are added to the
    write request as they're unlocked, but there may be a delay before
    ALL_QUEUED is set as the write subrequests may complete before we get
    there.
    
    If all the write subreqs have finished by the ALL_QUEUED point, no further
    events happen and the collection never happens, leaving the request
    hanging.
    
    Fix this by queuing the collector after setting ALL_QUEUED.  This is a bit
    heavy-handed and it may be sufficient to do it only if there are no extant
    subreqs.
    
    Also add a tracepoint to cross-reference both requests in a copy-to-request
    operation and add a trace to the netfs_rreq tracepoint to indicate the
    setting of ALL_QUEUED.
    
    Fixes: e2d46f2ec332 ("netfs: Change the read result collector to only use one work item")
    Reported-by: Max Kellermann <max.kellermann@ionos.com>
    Link: https://lore.kernel.org/r/CAKPOu+8z_ijTLHdiCYGU_Uk7yYD=shxyGLwfe-L7AV3DhebS3w@mail.gmail.com/
    Signed-off-by: David Howells <dhowells@redhat.com>
    cc: Paulo Alcantara <pc@manguebit.org>
    cc: Viacheslav Dubeyko <slava@dubeyko.com>
    cc: Alex Markuze <amarkuze@redhat.com>
    cc: Ilya Dryomov <idryomov@gmail.com>
    cc: netfs@lists.linux.dev
    cc: ceph-devel@vger.kernel.org
    cc: linux-fsdevel@vger.kernel.org
    cc: stable@vger.kernel.org

diff --git a/fs/netfs/read_pgpriv2.c b/fs/netfs/read_pgpriv2.c
index 080d2a6a51d9..889ff7954f8c 100644
--- a/fs/netfs/read_pgpriv2.c
+++ b/fs/netfs/read_pgpriv2.c
@@ -111,6 +111,7 @@ static struct netfs_io_request *netfs_pgpriv2_begin_copy_to_cache(
 		goto cancel_put;
 
 	__set_bit(NETFS_RREQ_OFFLOAD_COLLECTION, &creq->flags);
+	trace_netfs_copy2cache(rreq, creq);
 	trace_netfs_write(creq, netfs_write_trace_copy_to_cache);
 	netfs_stat(&netfs_n_wh_copy_to_cache);
 	rreq->copy_to_cache = creq;
@@ -155,6 +156,8 @@ void netfs_pgpriv2_end_copy_to_cache(struct netfs_io_request *rreq)
 	netfs_issue_write(creq, &creq->io_streams[1]);
 	smp_wmb(); /* Write lists before ALL_QUEUED. */
 	set_bit(NETFS_RREQ_ALL_QUEUED, &creq->flags);
+	trace_netfs_rreq(rreq, netfs_rreq_trace_end_copy_to_cache);
+	netfs_wake_collector(creq);
 
 	netfs_put_request(creq, netfs_rreq_trace_put_return);
 	creq->copy_to_cache = NULL;
diff --git a/include/trace/events/netfs.h b/include/trace/events/netfs.h
index 73e96ccbe830..64a382fbc31a 100644
--- a/include/trace/events/netfs.h
+++ b/include/trace/events/netfs.h
@@ -55,6 +55,7 @@
 	EM(netfs_rreq_trace_copy,		"COPY   ")	\
 	EM(netfs_rreq_trace_dirty,		"DIRTY  ")	\
 	EM(netfs_rreq_trace_done,		"DONE   ")	\
+	EM(netfs_rreq_trace_end_copy_to_cache,	"END-C2C")	\
 	EM(netfs_rreq_trace_free,		"FREE   ")	\
 	EM(netfs_rreq_trace_ki_complete,	"KI-CMPL")	\
 	EM(netfs_rreq_trace_recollect,		"RECLLCT")	\
@@ -559,6 +560,35 @@ TRACE_EVENT(netfs_write,
 		      __entry->start, __entry->start + __entry->len - 1)
 	    );
 
+TRACE_EVENT(netfs_copy2cache,
+	    TP_PROTO(const struct netfs_io_request *rreq,
+		     const struct netfs_io_request *creq),
+
+	    TP_ARGS(rreq, creq),
+
+	    TP_STRUCT__entry(
+		    __field(unsigned int,		rreq)
+		    __field(unsigned int,		creq)
+		    __field(unsigned int,		cookie)
+		    __field(unsigned int,		ino)
+			     ),
+
+	    TP_fast_assign(
+		    struct netfs_inode *__ctx = netfs_inode(rreq->inode);
+		    struct fscache_cookie *__cookie = netfs_i_cookie(__ctx);
+		    __entry->rreq	= rreq->debug_id;
+		    __entry->creq	= creq->debug_id;
+		    __entry->cookie	= __cookie ? __cookie->debug_id : 0;
+		    __entry->ino	= rreq->inode->i_ino;
+			   ),
+
+	    TP_printk("R=%08x CR=%08x c=%08x i=%x ",
+		      __entry->rreq,
+		      __entry->creq,
+		      __entry->cookie,
+		      __entry->ino)
+	    );
+
 TRACE_EVENT(netfs_collect,
 	    TP_PROTO(const struct netfs_io_request *wreq),