Skip to content
  1. Jun 23, 2015
  3. Jun 20, 2015
  5. Jun 05, 2015
    • Chuck Lever's avatar
      rpcrdma: Merge svcrdma and xprtrdma modules into one · ffe1f0df
      Chuck Lever authored 
      Bi-directional RPC support means code in svcrdma.ko invokes a bit of
code in xprtrdma.ko, and vice versa. To avoid loader/linker loops,
merge the server and client side modules together into a single
module.

When backchannel capabilities are added, the combined module will
register all needed transport capabilities so that Upper Layer
consumers automatically have everything needed to create a
bi-directional transport connection.

Module aliases are added for backwards compatibility with user
space, which still may expect svcrdma.ko or xprtrdma.ko to be
present.

This commit reverts commit 2e8c12e1

 ("xprtrdma: add separate
Kconfig options for NFSoRDMA client and server support") and
provides a single CONFIG option for enabling the new module.

Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
Signed-off-by: default avatarJ. Bruce Fields <bfields@redhat.com>
      ffe1f0df
    • Chuck Lever's avatar
      svcrdma: Add a separate "max data segs macro for svcrdma · 0380a3f3
      Chuck Lever authored 
      

The server and client maximum are architecturally independent.
Allow changing one without affecting the other.

Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
Signed-off-by: default avatarJ. Bruce Fields <bfields@redhat.com>
      0380a3f3
    • Chuck Lever's avatar
      svcrdma: Replace GFP_KERNEL in a loop with GFP_NOFAIL · b7e0b9a9
      Chuck Lever authored 
      

At the 2015 LSF/MM, it was requested that memory allocation
call sites that request GFP_KERNEL allocations in a loop should be
annotated with __GFP_NOFAIL.

Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
Signed-off-by: default avatarJ. Bruce Fields <bfields@redhat.com>
      b7e0b9a9
    • Chuck Lever's avatar
      svcrdma: Keep rpcrdma_msg fields in network byte-order · 30b7e246
      Chuck Lever authored 
      

Fields in struct rpcrdma_msg are __be32. Don't byte-swap these
fields when decoding RPC calls and then swap them back for the
reply. For the most part, they can be left alone.

Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
Signed-off-by: default avatarJ. Bruce Fields <bfields@redhat.com>
      30b7e246
    • Chuck Lever's avatar
      svcrdma: Fix byte-swapping in svc_rdma_sendto.c · 70747c25
      Chuck Lever authored 
      In send_write_chunks(), we have:

	for (xdr_off = rqstp->rq_res.head[0].iov_len, chunk_no = 0;
	     xfer_len && chunk_no < arg_ary->wc_nchunks;
	     chunk_no++) {
		 . . .
	}

Note that arg_ary->wc_nchunk is in network byte-order. For the
comparison to work correctly, both have to be in native byte-order.

In send_reply_chunks, we have:

	write_len = min(xfer_len, htonl(ch->rs_length));

xfer_len is in native byte-order, and ch->rs_length is in
network byte-order. be32_to_cpu() is the correct byte swap
for ch->rs_length.

As an additional clean up, replace ntohl() with be32_to_cpu() in
a few other places.

This appears to address a problem with large rsize hangs while
using PHYSICAL memory registration. I suspect that is the only
registration mode that uses more than one chunk element.

BugLink: https://bugzilla.linux-nfs.org/show_bug.cgi?id=248


Signed-off-by: default avatarChuck Lever <chuck.lever@oracle.com>
Signed-off-by: default avatarJ. Bruce Fields <bfields@redhat.com>
      70747c25
    • Kinglong Mee's avatar
      nfsd: Update callback sequnce id only CB_SEQUENCE success · 276f03e3
      Kinglong Mee authored 
      

When testing pnfs layout, nfsd got error NFS4ERR_SEQ_MISORDERED.
It is caused by nfs return NFS4ERR_DELAY before validate_seqid(),
don't update the sequnce id, but nfsd updates the sequnce id !!!

According to RFC5661 20.9.3,
" If CB_SEQUENCE returns an error, then the state of the slot
(sequence ID, cached reply) MUST NOT change. "

Signed-off-by: default avatarKinglong Mee <kinglongmee@gmail.com>
Signed-off-by: default avatarJ. Bruce Fields <bfields@redhat.com>
      276f03e3
    • Kinglong Mee's avatar
      nfsd: Reset cb_status in nfsd4_cb_prepare() at retrying · 4399396e
      Kinglong Mee authored 
      

nfsd enters a infinite loop and prints message every 10 seconds:

May 31 18:33:52 test-server kernel: Error sending entire callback!
May 31 18:34:01 test-server kernel: Error sending entire callback!

This is caused by a cb_layoutreturn getting error -10008
(NFS4ERR_DELAY), the client crashing, and then nfsd entering the
infinite loop:

  bc_sendto --> call_timeout --> nfsd4_cb_done --> nfsd4_cb_layout_done
  with error -10008 --> rpc_delay(task, HZ/100) --> bc_sendto ...

Reproduced using xfstests 074 with nfs client's kdump on,
CONFIG_DEFAULT_HUNG_TASK_TIMEOUT set, and client's blkmapd down:

1. nfs client's write operation will get the layout of file,
   and then send getdeviceinfo,
2. but layout segment is not recorded by client because blkmapd is down,
3. client writes data by sending WRITE to server,
4. nfs server recalls the layout of the file before WRITE,
5. network error causes the client reset the session and return NFS4ERR_DELAY,
6. so client's WRITE operation is waiting the reply.
   If the task hangs 120s, the client will crash.
7. so that, the next bc_sendto will fail with TIMEOUT,
   and cb_status is NFS4ERR_DELAY.

Signed-off-by: default avatarKinglong Mee <kinglongmee@gmail.com>
Signed-off-by: default avatarJ. Bruce Fields <bfields@redhat.com>
      4399396e
  7. Jun 04, 2015
  9. Jun 01, 2015
  11. May 29, 2015
  13. May 07, 2015
  15. May 05, 2015
  17. May 04, 2015
  19. May 03, 2015
    • Jan Kara's avatar
      ext4: fix growing of tiny filesystems · 2c869b26
      Jan Kara authored 
      

The estimate of necessary transaction credits in ext4_flex_group_add()
is too pessimistic. It reserves credit for sb, resize inode, and resize
inode dindirect block for each group added in a flex group although they
are always the same block and thus it is enough to account them only
once. Also the number of modified GDT block is overestimated since we
fit EXT4_DESC_PER_BLOCK(sb) descriptors in one block.

Make the estimation more precise. That reduces number of requested
credits enough that we can grow 20 MB filesystem (which has 1 MB
journal, 79 reserved GDT blocks, and flex group size 16 by default).

Signed-off-by: default avatarJan Kara <jack@suse.cz>
Signed-off-by: default avatarTheodore Ts'o <tytso@mit.edu>
Reviewed-by: default avatarEric Sandeen <sandeen@redhat.com>
      2c869b26