[linux-cifs-client] [PATCH 2/4] CIFS fixes for 2.4.x

[Sergey Vlasov]

This patch sets sk->allocation for the cifs socket to GFP_NOFS, so
the internal memory allocations in the socket layer will not cause
attempts to reenter the cifs code for writeout of dirty pages.

I have not been able to reproduce a problem specifically related to
these allocations, however, this looks like the right way to avoid
reentrancy problems.

There was a commented out assignment to sk->allocation in smb_send()
(with a nonexistent GFP_BUFFER value) - I have deleted it and put
the setting to ipv4_connect() after the socket is created (it needs
to be done only once per socket).


--- kernel-source-cifs-1.0.2/connect.c.sock-alloc	2004-04-08 10:50:50 +0400
+++ kernel-source-cifs-1.0.2/connect.c	2004-04-08 12:54:18 +0400
@@ -798,6 +798,7 @@ ipv4_connect(struct sockaddr_in *psin_se
 			return rc;
 		} else {
 		/* BB other socket options to set KEEPALIVE, NODELAY? */
+			(*csocket)->sk->allocation = GFP_NOFS;
 			cFYI(1,("Socket created"));
 		}
 	}
--- kernel-source-cifs-1.0.2/transport.c.sock-alloc	2004-02-05 05:55:30 +0300
+++ kernel-source-cifs-1.0.2/transport.c	2004-04-08 12:54:35 +0400
@@ -127,7 +127,6 @@ smb_send(struct socket *ssocket, struct 
 
 	if(ssocket == NULL)
 		return -ENOTSOCK; /* BB eventually add reconnect code here */
-/*  ssocket->sk->allocation = GFP_BUFFER; *//* BB is this spurious? */
 	iov.iov_base = smb_buffer;
 	iov.iov_len = smb_buf_length + 4;
 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.samba.org/archive/linux-cifs-client/attachments/20040408/15cd2cce/attachment.bin