- 09 Oct, 2007 40 commits
-
-
Chuck Lever authored
The purpose of an RPC ping (a NULL request) is to determine whether the remote end is operating and supports the RPC program and version of the request. If we do an RPC bind and the remote's rpcbind service says "this program or service isn't supported" then we have our answer already, and we should give up immediately. This is good for the kernel mount client, as it will cause the request to fail, and then allow an immediate retry with different options. Signed-off-by:
Chuck Lever <chuck.lever@oracle.com> Signed-off-by:
Trond Myklebust <Trond.Myklebust@netapp.com>
-
Chuck Lever authored
Add more new error code processing to the kernel's rpcbind client and to call_bind_status() to distinguish two cases: Case 1: the remote has replied that the program/version tuple is not registered (returns EACCES) Case 2: retry with a lesser rpcbind version (rpcb now returns EPFNOSUPPORT) This change allows more specific error processing for each of these two cases. We now fail case 2 instead of retrying... it's a server configuration error not to support even rpcbind version 2. And don't expose this new error code to user land -- convert it to EIO before failing the RPC. Signed-off-by:
-
Chuck Lever authored
Add new error code processing to the kernel's rpcbind client and to call_bind_status() to distinguish two cases: Case 1: the remote has replied that the program/version tuple is not registered (returns -EACCES) Case 2: another process is already in the middle of binding on this transport (now returns -EAGAIN) This change allows more specific retry processing for each of these two cases. Signed-off-by:
-
Chuck Lever authored
When a server returns a bad rpcbind reply, make rpcbind client recovery logic retry with an older protocol version. Older versions are more likely to work correctly. Signed-off-by:
-
Chuck Lever authored
Add better sanity checking of server replies to the GETVERSADDR reply decoder. Change the error return code: EIO is what other XDR decoding routines return if there is a failure while decoding. Signed-off-by:
-
Chuck Lever authored
/home/cel/linux/net/sunrpc/clnt.c: In function ‘rpc_show_tasks’: /home/cel/linux/net/sunrpc/clnt.c:1538: warning: signed and unsigned type in conditional expression This points out another case where a conditional expression returns a signed value in one arm and an unsigned value in the other. Signed-off-by:
-
Chuck Lever authored
Check the length of the passed-in server name before trying to print it in the log. Signed-off-by:
-
Chuck Lever authored
Noticed by Tom Talpey <tmt@netapp.com>: OBTW, there's a nit on that memcpy, too. The r_addr is an array, so memcpy(&map->r_addr is passing the address of the array as a char **. It's the same as map->r_addr, but technically the wrong type. Signed-off-by:
-
Chuck Lever authored
/home/cel/linux/net/sunrpc/clnt.c: In function ‘rpc_bind_new_program’: /home/cel/linux/net/sunrpc/clnt.c:445: warning: comparison between signed and unsigned RPC version numbers are u32, not int. Signed-off-by:
-
Chuck Lever authored
Remove one of two identical dprintk's that occur when an RPC client is created. Signed-off-by:
-
Chuck Lever authored
Fix some problems with rpcbind v3 and v4 queries from the in-kernel rpcbind client: 1. The r_addr argument must be a full universal address, not just an IP address, and 2. The universal address in r_addr is the address of the remote rpcbind server, not the RPC service being requested This addresses bugzilla.kernel.org report 8891 for 2.6.23-rc and greater. In addition, if the rpcbind client is unable to start the rpcbind request, make sure not to leak the xprt. Signed-off-by: -
Chuck Lever authored
"Universal addresses" are a string representation of an IP address and port. They are described fully in RFC 3530, section 2.2. Add support for generating them in the RPC client's socket transport module. Signed-off-by: -
Chuck Lever authored
Signed-off-by:
-
Chuck Lever authored
Signed-off-by:
-
Chuck Lever authored
Finalize support for setting up RPC client transports to remote RPC services addressed via IPv6. Based on work done by Gilles Quillard at Bull Open Source. Signed-off-by:
-
Chuck Lever authored
Clone separate connect worker functions for connecting AF_INET6 sockets. Signed-off-by:
-
Chuck Lever authored
Prepare for introduction of IPv6 versions of same. Signed-off-by:
-
Chuck Lever authored
Finishing a socket connect is the same for IPv4 and IPv6, so split it out into a helper. Signed-off-by:
-
Chuck Lever authored
Clone xs_bindresvport into two functions, one that can handle IPv4 addresses, and one that can handle IPv6 addresses. Signed-off-by:
-
Chuck Lever authored
Prepare for introduction of IPv6-specific socket bind function. Signed-off-by:
-
Chuck Lever authored
We could clone xs_set_port, but this is easier overall. Signed-off-by:
-
Chuck Lever authored
Prepare for adding IPv6 support to the RPC client by adding IPv6 capabilities to rpcbind. Note that this is support on the query side only; registering IPv6 addresses with the local portmapper will come later. Note we have to take care not to fall back to using version 2 of the rpcbind protocol if we're dealing with IPv6 address. Version 2 doesn't support IPv6 at all. Signed-off-by:
-
Chuck Lever authored
Clone xs_format_ipv4_peer_addresses into an IPv6 version. Signed-off-by:
-
Chuck Lever authored
Prepare to add an IPv6 version of xs_format_peer_addresses by renaming it to xs_format_ipv4_peer_addresses. Signed-off-by:
-
Chuck Lever authored
Add support for the NFS client's need to export volume information with IP addresses formatted in hex instead of decimal. This isn't used yet, but subsequent patches (not in this series) will change the NFS client to use this functionality. Signed-off-by:
-
Chuck Lever authored
Use more generic logic to free buffers holding formatted addresses. This makes it less likely a bug will be introduced when adding additional buffer types in xs_format_peer_address(). Signed-off-by:
-
Chuck Lever authored
include/linux/kernel.h gives us some nice macros for formatting IP addresses. Use them. Signed-off-by:
-
Chuck Lever authored
Signed-off-by:
-
Chuck Lever authored
Signed-off-by:
-
Jeff Layton authored
A minor thing, but useful when working with a server with multiple addrs. This looks like it might also be necessary if Miklos' effort to eliminate /etc/mtab ever comes to fruition. When displaying mount options in /proc/mounts, the kernel prints "addr=hostname". This info is redundant since we already have the hostname displayed as part of the "device" section of the mount. This patch changes it to display the IP address to which the socket is connected. Signed-off-by:
Jeff Layton <jlayton@redhat.com> Signed-off-by:
-
Christoph Hellwig authored
no need to set up foo-objs these days. Signed-off-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
-
Fabio Olive Leite authored
I would like to discuss the idea that the current checks for attribute timeout using time_after are inadequate for 32bit architectures, since time_after works correctly only when the two timestamps being compared are within 2^31 jiffies of each other. The signed overflow caused by comparing values more than 2^31 jiffies apart will flip the result, causing incorrect assumptions of validity. 2^31 jiffies is a fairly large period of time (~25 days) when compared to the lifetime of most kernel data structures, but for long lived NFS mounts that can sit idle for months (think that for some reason autofs cannot be used), it is easy to compare inode attribute timestamps with very disparate or even bogus values (as in when jiffies have wrapped many times, where the comparison doesn't even make sense). Currently the code tests for attribute timeout by simply adding the desired amount of jiffies to the stored timestamp and comparing that with the current timestamp of obtained attribute data with time_after. This is incorrect, as it returns true for the desired timeout period and another full 2^31 range of jiffies. In testing with artificial jumps (several small jumps, not one big crank) of the jiffies I was able to reproduce a problem found in a server with very long lived NFS mounts, where attributes would not be refreshed even after touching files and directories in the server: Initial uptime: 03:42:01 up 6 min, 0 users, load average: 0.01, 0.12, 0.07 NFS volume is mounted and time is advanced: 03:38:09 up 25 days, 2 min, 0 users, load average: 1.22, 1.05, 1.08 # ls -l /local/A/foo/bar /nfs/A/foo/bar -rw-r--r-- 1 root root 0 Dec 17 03:38 /local/A/foo/bar -rw-r--r-- 1 root root 0 Nov 22 00:36 /nfs/A/foo/bar # touch /local/A/foo/bar # ls -l /local/A/foo/bar /nfs/A/foo/bar -rw-r--r-- 1 root root 0 Dec 17 03:47 /local/A/foo/bar -rw-r--r-- 1 root root 0 Nov 22 00:36 /nfs/A/foo/bar We can see the local mtime is updated, but the NFS mount still shows the old value. The patch below makes it work: Initial setup... 07:11:02 up 25 days, 1 min, 0 users, load average: 0.15, 0.03, 0.04 # ls -l /local/A/foo/bar /nfs/A/foo/bar -rw-r--r-- 1 root root 0 Jan 11 07:11 /local/A/foo/bar -rw-r--r-- 1 root root 0 Jan 11 07:11 /nfs/A/foo/bar # touch /local/A/foo/bar # ls -l /local/A/foo/bar /nfs/A/foo/bar -rw-r--r-- 1 root root 0 Jan 11 07:14 /local/A/foo/bar -rw-r--r-- 1 root root 0 Jan 11 07:14 /nfs/A/foo/bar Signed-off-by:
Fabio Olive Leite <fleite@redhat.com> Signed-off-by:
-
Peter Staubach authored
Hi. Attached is a patch to modify the NFS client code to support 64 bit ino's, as appropriate for the system and the NFS protocol version. The code basically just expand the NFS interfaces for routines which handle ino's from using ino_t to u64 and then uses the fileid in the nfs_inode instead of i_ino in the inode. The code paths that were updated are in the getattr method and the readdir methods. This should be no real change on 64 bit platforms. Since the ino_t is an unsigned long, it would already be 64 bits wide. Thanx... ps Signed-off-by:Peter Staubach <staubach@redhat.com> Signed-off-by:
-
Trond Myklebust authored
This will allow rpc.gssd to use inotify instead of dnotify in order to locate new rpc upcall pipes. This also requires the exporting of __audit_inode_child(), which is used by fsnotify_create() and fsnotify_mkdir(). Ccing David Woodhouse. Cc: David Woodhouse <dwmw2@infradead.org> Signed-off-by: -
Trond Myklebust authored
This helps prevent huge queues of background writes from building up whenever the server runs out of disk or quota space, or if someone changes the file access modes behind our backs. Signed-off-by: -
Trond Myklebust authored
Schedule writes using WB_SYNC_NONE first, then come back for a second pass using WB_SYNC_ALL. Signed-off-by: -
Trond Myklebust authored
The only user of nfs_sync_mapping_range() is nfs_getattr(), which uses it to flush out the entire inode without sending a commit. We therefore replace nfs_sync_mapping_range with a more appropriate helper. Signed-off-by: -
Trond Myklebust authored
The only user of this field was NFS. Signed-off-by: -
Trond Myklebust authored
Just call write_cache_pages directly instead of hacking the writeback control structure in order to find out if we were called from writepages() or directly from the VM. Signed-off-by: -
Trond Myklebust authored
The addition of nfs_page_mkwrite means that We should no longer need to create requests inside nfs_writepage() Signed-off-by:
-
