SLAB_NO_REAP is documented as an option that will cause this slab not to be
reaped under memory pressure.  However, that is not what happens.  The only
thing that SLAB_NO_REAP controls at the moment is the reclaim of the unused
slab elements that were allocated in batch in cache_reap().  Cache_reap()
is run every few seconds independently of memory pressure.

Could we remove the whole thing?  Its only used by three slabs anyways and
I cannot find a reason for having this option.

There is an additional problem with SLAB_NO_REAP.  If set then the recovery
of objects from alien caches is switched off.  Objects not freed on the
same node where they were initially allocated will only be reused if a
certain amount of objects accumulates from one alien node (not very likely)
or if the cache is explicitly shrunk.  (Strangely __cache_shrink does not
check for SLAB_NO_REAP)

Getting rid of SLAB_NO_REAP fixes the problems with alien cache freeing.
Signed-off-by: NChristoph Lameter <clameter@sgi.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Manfred Spraul <manfred@colorfullife.com>
Cc: Mark Fasheh <mark.fasheh@oracle.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

>From ogerlitz@voltaire.com:

mgmtpool shoild be frees in immdata_alloc_fail label.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

>From erezz@voltaire.com:

We are still in ISCSI_STATE_FREE state at create time. The addition
of the first connection puts us in ISCSI_STATE_LOGGED_IN.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

>From erezz@voltaire.com:

rm conn->lock since it is not used anymore. The dataqueue is protected
by the session lock and xmitmutex.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From:
michaelc@cs.wisc.edu
fujita.tomonori@lab.ntt.co.jp
da-x@monatomic.org

and err path fixup from:
ogerlitz@voltaire.com

This patch cleans up that interface by having the lld and class
pass a iscsi_cls_session or iscsi_cls_conn between each other when
the function is used by HW and SW iscsi llds. This way the lld
does not have to remember if it has to send a handle or pointer
and a handle or pointer to connection, session or host.

This also has the class verify the session handle that gets passed from
userspace instead of using the pointer passed into the kernel directly.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Remove the "inline" keyword from a bunch of big functions in the kernel with
the goal of shrinking it by 30kb to 40kb
Signed-off-by: NArjan van de Ven <arjan@infradead.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Acked-by: NJeff Garzik <jgarzik@pobox.com>
Signed-off-by: NAndrew Morton <akpm@osdl.org>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

From: FUJITA Tomonori <tomof@acm.org> and zhenyu.z.wang@intel.com:

We cannot handle filesystems like XFS becuase of the pages they
are sending us. We had thought page_count could be used to
work around this, but the correct test is for PageSlab.

The proper solution is to figure out what type of pages
filesystems can use so we do not have to add tests like
this or handle it in the block layer for all network block drivers
but the issue still has not been resolved on fs-devel
so we are sending this patch as a temporary fix.

This is last patch just in case it is Nakd with the explanation
that we need to push the correct fix through fs-devel, mm
or the block layer. The rest of the patchset can live without
the patch, but the driver will not work with filesystems like
XFS.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

When we run the xmit code from queuecomand the stack trace
gets too deep. The patch runs the xmit code from the scsi_host
work queue. This fixes 4k stack and xfs support and should
fix the st and sg stack usage bugs.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

This is the second version of the patch to address Christoph's comments.
Instead of doing the lib, I just kept everything in scsi_trnapsort_iscsi.c
like the FC and SPI class. This was becuase the driver model and sysfs
class is tied to the session and connection setup so separating did not
buy very much at this time.

The reason for this patch was becuase HW iscsi LLDs like qla4xxx cannot
use the iscsi class becuase the scsi_host was tied to the interface and
class code. This patch just seperates the session from scsi host so
that LLDs that allocate the host per some resource like pci device
can still use the class.

This is also fixes a couple refcount bugs that can be triggered
when users have a sysfs file open, close the session, then
read or write to the file.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From Mike Christie <michaelc@cs.wisc.edu> and FUJITA Tomonori <tomof@acm.org>:

We cannot use page_address becuase some pages could be highmem.
Instead, we can use sock_no_sendpage which does kmap for us.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Users can write to a page while we are sending it and making
digest calculations. This ends up causing us to retry the command
when a digest error is later reported. By using sock_no_sendpage
when data digests are calculated we can avoid a lot of (not all but it
helps) the retries becuase sock_no_sendpage is not zero copy.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

We should be taking the host_lock instead of the conn lock when
checking host_busy.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

We need to check the ISCSI_FLAG_DATA_* flags.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Remove extra whitespaces.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

the scsi layer is using semaphores in a mutex way, this patch converts
these into using mutexes instead
Signed-off-by: NArjan van de Ven <arjan@infradead.org>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NLinus Torvalds <torvalds@osdl.org>

From Wang Zhenyu:

check header digest for cmd and mgmt tasks
Signed-off-by: NWang Zhenyu <zhenyu.z.wang@intel.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From Wang Zhenyu:

High queue depth was a problem for some targets so make queue_depth adjustable

From Mike Christie

Make default queue_depth a little lower
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From Wang Zhenyu:

data digest fix (the bug caused data corruption w/Wasabi StorageBuilder target)
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

from Wang Zhenyu:

Must check SCSI CMD and R2T response according to the spec
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From tomof@acm.org:

There is one more issue about Equallogic systems. They send
re-direction info with FIN. I think that the kernel module needs to
let iscsid to read data from the socket before killing it.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

Must check only valid opcode bits.
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: michaelc@cs.wisc.edu

I have a bad memory. I cannot remember what versions are which,
so add a module version to help.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: zhenyu.z.wang@intel.com

This add check to NOOP_IN's ttt, when it's ~0UL we should not send
NOOP_OUT by spec (plus some cleanup).
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: tomof@acm.org

Fix AHS Length
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: michaelc@cs.wisc.edu

Cleanup some iscsi_proto defs, add some missing values, and
fix some defs.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: zhenyu.z.wang@intel.com

Fix oops from nonlinear skb usage.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: zhenyu.z.wang@intel.com

Delay the head digest update until xmit time, like data digest update.
[To make things cleaner and avoid prempt bug]
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

From: tomof@acm.org

I'm not sure about this. I don't think that NODELAY option hurts
performance. However, open-iscsi does not use MSG_MORE properly with
sendpage, so NODELAY option hurts the open-iscsi performance.

I've attached a patch to fix NODELAY and MSG_MORE problems and the
write performance results with disktest.

I use Opteron boxes connected directly, Chelsio NICs, 1500-byte MTU,
64 KB I/O size, and the iSCSI parameters on open-iscsi web site.

With only NODELAY fix, the performance drops, as you said. On the
other hand, NODELAY and MSG_MORE fixes improve the performance
overall.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

drivers/scsi/iscsi_tcp.c, iscsi data path.
Signed-off-by: NAlex Aizman <itn780@yahoo.com>
Signed-off-by: NDmitry Yusupov <dmitry_yus@yahoo.com>
Signed-off-by: NMike Christie <michaelc@cs.wisc.edu>
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>