From edf1ae403896cb7750800508b14996ba6be39a53 Mon Sep 17 00:00:00 2001 From: Steve French Date: Wed, 29 Oct 2008 00:47:57 +0000 Subject: [CIFS] Reduce number of socket retries in large write path CIFS in some heavy stress conditions cifs could get EAGAIN repeatedly in smb_send2 which led to repeated retries and eventually failure of large writes which could lead to data corruption. There are three changes that were suggested by various network developers: 1) convert cifs from non-blocking to blocking tcp sendmsg (we left in the retry on failure) 2) change cifs to not set sendbuf and rcvbuf size for the socket (let tcp autotune the buffer sizes since that works much better in the TCP stack now) 3) if we have a partial frame sent in smb_send2, mark the tcp session as invalid (close the socket and reconnect) so we do not corrupt the remaining part of the SMB with the beginning of the next SMB. This does not appear to hurt performance measurably and has been run in various scenarios, but it definately removes a corruption that we were seeing in some high stress test cases. Acked-by: Shirish Pargaonkar Signed-off-by: Steve French --- fs/cifs/CHANGES | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'fs/cifs/CHANGES') diff --git a/fs/cifs/CHANGES b/fs/cifs/CHANGES index 8f528ea24c48..8855331b2fba 100644 --- a/fs/cifs/CHANGES +++ b/fs/cifs/CHANGES @@ -4,7 +4,11 @@ Various fixes to make delete of open files behavior more predictable (when delete of an open file fails we mark the file as "delete-on-close" in a way that more servers accept, but only if we can first rename the file to a temporary name). Add experimental support for more safely -handling fcntl(F_SETLEASE). +handling fcntl(F_SETLEASE). Convert cifs to using blocking tcp +sends, and also let tcp autotune the socket send and receive buffers. +This reduces the number of EAGAIN errors returned by TCP/IP in +high stress workloads (and the number of retries on socket writes +when sending large SMBWriteX requests). Version 1.54 ------------ -- cgit v1.2.3 From bfb59820ee46616a7bdb4af6b8f7e109646de6ec Mon Sep 17 00:00:00 2001 From: Steve French Date: Tue, 18 Nov 2008 16:33:48 +0000 Subject: [CIFS] fix check for dead tcon in smb_init This was recently changed to check for need_reconnect, but should actually be a check for a tidStatus of CifsExiting. Signed-off-by: Jeff Layton Signed-off-by: Steve French --- fs/cifs/CHANGES | 6 +++++- fs/cifs/cifssmb.c | 2 +- 2 files changed, 6 insertions(+), 2 deletions(-) (limited to 'fs/cifs/CHANGES') diff --git a/fs/cifs/CHANGES b/fs/cifs/CHANGES index 8855331b2fba..e078b7aea143 100644 --- a/fs/cifs/CHANGES +++ b/fs/cifs/CHANGES @@ -8,7 +8,11 @@ handling fcntl(F_SETLEASE). Convert cifs to using blocking tcp sends, and also let tcp autotune the socket send and receive buffers. This reduces the number of EAGAIN errors returned by TCP/IP in high stress workloads (and the number of retries on socket writes -when sending large SMBWriteX requests). +when sending large SMBWriteX requests). Fix case in which a portion of +data can in some cases not get written to the file on the server before the +file is closed. Fix DFS parsing to properly handle path consumed field, +and to handle certain codepage conversions better. Fix mount and +umount race that can cause oops in mount or umount or reconnect. Version 1.54 ------------ diff --git a/fs/cifs/cifssmb.c b/fs/cifs/cifssmb.c index bdda46dd435a..2af8626ced43 100644 --- a/fs/cifs/cifssmb.c +++ b/fs/cifs/cifssmb.c @@ -295,7 +295,7 @@ smb_init(int smb_command, int wct, struct cifsTconInfo *tcon, check for tcp and smb session status done differently for those three - in the calling routine */ if (tcon) { - if (tcon->need_reconnect) { + if (tcon->tidStatus == CifsExiting) { /* only tree disconnect, open, and write, (and ulogoff which does not have tcon) are allowed as we start force umount */ -- cgit v1.2.3