Right now SCSI and others do their own command timeout handling.
Move those bits to the block layer.
Instead of having a timer per command, we try to be a bit more clever
and simply have one per-queue. This avoids the overhead of having to
tear down and setup a timer for each command, so it will result in a lot
less timer fiddling.
Signed-off-by: Mike Anderson <andmike@linux.vnet.ibm.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
When running ibmvscsi in a shared memory partition, it must provide
a default value for the amount of DMA resources it will need in order to
perform reasonably well. This was being calculated in sectors rather than
bytes, as it should. This patch fixes this.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Add per-device dma_mapping_ops support for CONFIG_X86_64 as POWER
architecture does:
This enables us to cleanly fix the Calgary IOMMU issue that some devices
are not behind the IOMMU (http://lkml.org/lkml/2008/5/8/423).
I think that per-device dma_mapping_ops support would be also helpful for
KVM people to support PCI passthrough but Andi thinks that this makes it
difficult to support the PCI passthrough (see the above thread). So I
CC'ed this to KVM camp. Comments are appreciated.
A pointer to dma_mapping_ops to struct dev_archdata is added. If the
pointer is non NULL, DMA operations in asm/dma-mapping.h use it. If it's
NULL, the system-wide dma_ops pointer is used as before.
If it's useful for KVM people, I plan to implement a mechanism to register
a hook called when a new pci (or dma capable) device is created (it works
with hot plugging). It enables IOMMUs to set up an appropriate
dma_mapping_ops per device.
The major obstacle is that dma_mapping_error doesn't take a pointer to the
device unlike other DMA operations. So x86 can't have dma_mapping_ops per
device. Note all the POWER IOMMUs use the same dma_mapping_error function
so this is not a problem for POWER but x86 IOMMUs use different
dma_mapping_error functions.
The first patch adds the device argument to dma_mapping_error. The patch
is trivial but large since it touches lots of drivers and dma-mapping.h in
all the architecture.
This patch:
dma_mapping_error() doesn't take a pointer to the device unlike other DMA
operations. So we can't have dma_mapping_ops per device.
Note that POWER already has dma_mapping_ops per device but all the POWER
IOMMUs use the same dma_mapping_error function. x86 IOMMUs use device
argument.
[akpm@linux-foundation.org: fix sge]
[akpm@linux-foundation.org: fix svc_rdma]
[akpm@linux-foundation.org: build fix]
[akpm@linux-foundation.org: fix bnx2x]
[akpm@linux-foundation.org: fix s2io]
[akpm@linux-foundation.org: fix pasemi_mac]
[akpm@linux-foundation.org: fix sdhci]
[akpm@linux-foundation.org: build fix]
[akpm@linux-foundation.org: fix sparc]
[akpm@linux-foundation.org: fix ibmvscsi]
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Muli Ben-Yehuda <muli@il.ibm.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Avi Kivity <avi@qumranet.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Enable the driver to function in a Cooperative Memory Overcommitment (CMO)
environment.
The following changes are made to enable the driver for CMO:
* DMA mapping errors will not result in error messages if entitlement has
been exceeded and resources were not available.
* The driver has a get_desired_dma function defined to function
in a CMO environment. It will indicate how much IO memory it would like
to function.
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Acked by: Brian King <brking@linux.vnet.ibm.com>
Acked-by: Paul Mackerras <paulus@samba.org>
Acked-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Some versions of the Virtual I/O Server on Power
return 0x99 in the non-SCSI error status field as success,
rather than 0. This fixes the ibmvscsi driver to treat this
response as success.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
- struct scsi_cmnd had a 16 bytes command buffer of its own.
This is an unnecessary duplication and copy of request's
cmd. It is probably left overs from the time that scsi_cmnd
could function without a request attached. So clean that up.
- Once above is done, few places, apart from scsi-ml, needed
adjustments due to changing the data type of scsi_cmnd->cmnd.
- Lots of drivers still use MAX_COMMAND_SIZE. So I have left
that #define but equate it to BLK_MAX_CDB. The way I see it
and is reflected in the patch below is.
MAX_COMMAND_SIZE - means: The longest fixed-length (*) SCSI CDB
as per the SCSI standard and is not related
to the implementation.
BLK_MAX_CDB. - The allocated space at the request level
- I have audit all ISA drivers and made sure none use ->cmnd in a DMA
Operation. Same audit was done by Andi Kleen.
(*)fixed-length here means commands that their size can be determined
by their opcode and the CDB does not carry a length specifier, (unlike
the VARIABLE_LENGTH_CMD(0x7f) command). This is actually not exactly
true and the SCSI standard also defines extended commands and
vendor specific commands that can be bigger than 16 bytes. The kernel
will support these using the same infrastructure used for VARLEN CDB's.
So in effect MAX_COMMAND_SIZE means the maximum size command
scsi-ml supports without specifying a cmd_len by ULD's
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Adds support to the ibmvscsi driver to handle non SCSI error
status. This is needed to support some new VIOS enhancements.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Santiago Leon <santil@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
It's big, but there doesn't seem to be a way to split it up smaller...
Signed-off-by: Tony Jones <tonyj@suse.de>
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Cc: Roland Dreier <rolandd@cisco.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Hal Rosenstock <hal.rosenstock@gmail.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
With the sg table code, every SCSI driver is now either chain capable
or broken (or has sg_tablesize set so chaining is never activated), so
there's no need to have a check in the host template.
Also tidy up the code by moving the scatterlist size defines into the
SCSI includes and permit the last entry of the scatterlist pools not
to be a power of two.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
CRQ send errors that return with H_CLOSED should return with
SCSI_MLQUEUE_HOST_BUSY until firmware alerts the client of a CRQ
transport event. The transport event will either reinitialize and
requeue the requests or fail and return IO with DID_ERROR.
To avoid failing the eh_* functions while re-attaching to the server
adapter this will retry for a period of time while ibmvscsi_send_srp_event
returns SCSI_MLQUEUE_HOST_BUSY.
In ibmvscsi_eh_abort_handler() the loop includes the search of the
event list. The lock on the hostdata is dropped while waiting to try
again after failing ibmvscsi_send_srp_event. The event could have been
purged if a login was in progress when the function was called.
In ibmvscsi_eh_device_reset_handler() the loop includes the call to
get_event_struct() because a failing call to ibmvscsi_send_srp_event()
will have freed the event struct.
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Set the default command timeout for ibmvscsi disks to 60 seconds
to ensure we don't prematurely timeout commands. This fixes a problem
seen where the default 30 seconds was not long enough due to
congestion on the server.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
By setting the request_limit in send_srp_login to 1 we allowed login
requests to be sent to the server adapter. If this was not an initial
login, but was a login after a disconnect with the server, other I/O
requests could attempt to be processed before the login occured. These
I/O requests would fail, sometimes resulting in filesystems getting
marked read-only.
To address this we can set the request_limit to 0 while doing the login
and add an exception where login requests, along with task management
events, are always passed to the server.
There is a case where the request_limit had already reached 0 would result
in all events being sent rather than returning SCSI_MLQUEUE_HOST_BUSY; this
has also been fixed by this patch.
Signed-off-by: Robert Jennings <rcj@linux.vnet.ibm.com>
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This option is true if a low-level driver can support sg
chaining. This will be removed eventually when all the drivers are
converted to support sg chaining. q->max_phys_segments is set to
SCSI_MAX_SG_SEGMENTS if false.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
If you build a multiplatform kernel for iSeries and pSeries, with
ibmvscsic support, the resulting client doesn't work on iSeries.
This fixes that, using the appropriate low-level operations
for the machine detected at runtime.
[jejb: fixed up rejections around the srp transport patch]
Signed-off-by: David Woodhouse <dwmw2@infradead.org>
Acked by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
This adds a 'roles' attribute to rport like transport_fc. The role can
be initiator or target. That is, the initiator driver creates target
remote ports and the target driver creates initiator remote ports.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
This converts ibmvscsi to use the srp transport class.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Brian King <brking@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
No need to check use_sg since sg_tablesize is set appropriately in the
scsi host template.
Brian King's patch (2a7309372f) did this
cleanup but the data buffer accessors patch (written before the patch
and merged after it) restored the check.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
- remove the unnecessary map_single path.
- convert to use the new accessors for the sg lists and the
parameters.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Santiago Leon <santil@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The viosrp_crq timeout field is in seconds.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Santiago Leon <santil@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Since it is completely possible for scsi core to call
a LLDD's eh_abort function after the command has completed,
fix ibmvscsi to return SUCCESS if this is the case.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Fix a couple locking bugs discovered during code inspection.
ibmvscsi_send_srp_event needs to be called with the host lock
held. This patch fixes a couple paths in the code where this
wasn't true.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Adds an eh_host_reset_handler to ibmvscsi which resets the connection
to the vscsi server. This patch also adds a timer to internally
issues commands to prevent client hangs in the case of a misbehaving
server. Tested by modifying the VIOS such that it would occasionally
drop one or more request in sequence.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Converts ibmvscsi to use dev_printk and friends to simplify
debugging. ibmvscsi adapter initialization now looks like this:
ibmvscsi 30000005: SRP_VERSION: 16.a
ibmvscsi 30000005: partner initialization complete
ibmvscsi 30000005: sent SRP login
ibmvscsi 30000005: SRP_LOGIN succeeded
Additionally, this patch adds the logging of a couple return codes in
a couple logs.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Since sg_tablesize is set appropriately in the scsi host template,
remove the unnecessary check to make sure it is not exceeded
following the dma_map_sg call.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Adds support for a changeable queue depth to ibmvscsi.
Signed-off-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Fixed the kernel-doc comment for ibmvscsi_slave_configure. Thanks to
Randy Dunlap for pointing this out.
Adding a slave_configure function for the driver. Now the disks can be
restarted by the scsi mid-layer when the are disconnected and reconnected.
Signed-off-by: "Robert Jennings" <rcj@linux.vnet.ibm.com>
Signed-off-by: "Santiago Leon" <santil@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The request limit calculations used previously on the client failed to
mirror the state of the server. Additionally, when a value < 3 was provided
there could be problems setting can_queue and handling abort and reset
commands.
Signed-off-by: "Robert Jennings" <rcj@linux.vnet.ibm.com>
Signed-off-by: Santiago Leon <santil@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The "ibmvscsi: treat busy and error conditions separately" patch
submitted by Dave Boutcher back in June incorrectly reenables the CRQ.
The broken logic causes the adapter to get disabled if the CRQ
connection happens to close temporarily. This patch "fixes that
obviously wrong logic check" (Dave's words).
Signed-off-by: Santiago Leon <santil@us.ibm.com>
Signed-off-by: David Boutcher <sleddog@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
This patch fixes a condition where ibmvscsi treats a transport error as a
"busy" condition, so no errors were returned to the scsi mid-layer.
In a RAID environment this means that I/O hung rather than failing
over.
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Convert kmalloc + memset to kcalloc in ibmvscsi
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Acked-by: Dave Boutcher <sleddog@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
There is a window where we can be re-enabling an adapter, but
still allow SCSI commands to be sent to the target. This fix
sets our window (request_limit) to -1 as soon as we know the
adapter is being reenabled, and closes a very teeny tiny
window where we could set the window back to 1 before we
grab a lock.
Signed-off-by: Dave Boutcher <sleddog@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
New versions of the Power5 firmware can send a "re-enable" message to
the virtual scsi adapter. This fix makes us handle the message
correctly. Without it, the driver goes catatonic and the system crashes
unpleasantly.
Signed-off-by: Dave Boutcher <sleddog@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Just set the name field directly in the device_driver structure
contained in the vio_driver struct.
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Linda Xie ever so gently pointed out that she had a patch
to preserve compatibility with older SLES targets, and I told
her we didn't need to push it to mainline.
This patch explicitly checks the version of the IBMVSCSI target
and ensures that large scatterlists are not sent to older
targets.
Signed-off-by: Linda Xie <lxie@us.ibm.com>
Signed-off-by: Dave Boutcher <boutcher@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
The maximum size of a scatter-gather list that the current IBM VSCSI
Client can handle is 10. This patch adds large scatter-gather support
to the client so that it is capable of handling up to SG_ALL(255)
number of requests in the scatter-gather list.
Signed-off-by: Linda Xie <lxie@us.ibm.com>
Acked by: Dave C Boutcher <sleddog@us.ibm.com>
Rejections fixed up and
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
This patch fixes a long term borkenness in
ibmvscsi where we were using the wrong timeout
field from the scsi command (and using the
wrong units.) Now broken by the fact that the
scsi_cmnd timeout field is gone entirely.
This only worked before because all the SCSI
targets assumed that 0 was default.
Signed-off-by: Dave Boutcher <boutcher@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
With the removal of the spinlocking around eh calls, we need to add a
little more locking back in, otherwise we do some naked list
manipulation.
Signed-off-by: Dave Boutcher <boutcher@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Fix the problem in IBM VSCSI Client where the client doesn't send the
information which is expected by the server.
Signed-off-by: Linda Xie <lxie@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
