The MLX transport requires two extra gather entries for sends (one for
the header and one for the checksum at the end, as the comment says).
However the code checked that max_recv_sge was not too big, instead of
checking max_send_sge as it should have. Fix the code to check the
correct condition.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Exactly when the catastrophic error polling timer function runs is not
important, so use round_jiffies() to save unnecessary wakeups.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Since we use del_timer_sync() anyway, there's no need for an
additional flag to tell the timer not to rearm.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Commit b18aad71 ("IB/mthca: Fix RESET to ERROR transition") added some
extra code to handle a QP state transition from RESET to ERROR.
However, the latest 1.2.1 version of the IB spec has clarified that
this transition is actually not allowed, so we can remove this extra
code again.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This patch adds support for the IB "base memory management extension"
(BMME) and the equivalent iWARP operations (which the iWARP verbs
mandates all devices must implement). The new operations are:
- Allocate an ib_mr for use in fast register work requests.
- Allocate/free a physical buffer lists for use in fast register work
requests. This allows device drivers to allocate this memory as
needed for use in posting send requests (eg via dma_alloc_coherent).
- New send queue work requests:
* send with remote invalidate
* fast register memory region
* local invalidate memory region
* RDMA read with invalidate local memory region (iWARP only)
Consumer interface details:
- A new device capability flag IB_DEVICE_MEM_MGT_EXTENSIONS is added
to indicate device support for these features.
- New send work request opcodes IB_WR_FAST_REG_MR, IB_WR_LOCAL_INV,
IB_WR_RDMA_READ_WITH_INV are added.
- A new consumer API function, ib_alloc_mr() is added to allocate
fast register memory regions.
- New consumer API functions, ib_alloc_fast_reg_page_list() and
ib_free_fast_reg_page_list() are added to allocate and free
device-specific memory for fast registration page lists.
- A new consumer API function, ib_update_fast_reg_key(), is added to
allow the key portion of the R_Key and L_Key of a fast registration
MR to be updated. Consumers call this if desired before posting
a IB_WR_FAST_REG_MR work request.
Consumers can use this as follows:
- MR is allocated with ib_alloc_mr().
- Page list memory is allocated with ib_alloc_fast_reg_page_list().
- MR R_Key/L_Key "key" field is updated with ib_update_fast_reg_key().
- MR made VALID and bound to a specific page list via
ib_post_send(IB_WR_FAST_REG_MR)
- MR made INVALID via ib_post_send(IB_WR_LOCAL_INV),
ib_post_send(IB_WR_RDMA_READ_WITH_INV) or an incoming send with
invalidate operation.
- MR is deallocated with ib_dereg_mr()
- page lists dealloced via ib_free_fast_reg_page_list().
Applications can allocate a fast register MR once, and then can
repeatedly bind the MR to different physical block lists (PBLs) via
posting work requests to a send queue (SQ). For each outstanding
MR-to-PBL binding in the SQ pipe, a fast_reg_page_list needs to be
allocated (the fast_reg_page_list is owned by the low-level driver
from the consumer posting a work request until the request completes).
Thus pipelining can be achieved while still allowing device-specific
page_list processing.
The 32-bit fast register memory key/STag is composed of a 24-bit index
and an 8-bit key. The application can change the key each time it
fast registers thus allowing more control over the peer's use of the
key/STag (ie it can effectively be changed each time the rkey is
rebound to a page list).
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Current memfree FW has a bug which in some cases, assumes that ICM
pages passed to it are cleared. This patch uses __GFP_ZERO to
allocate all ICM pages passed to the FW. Once firmware with a fix is
released, we can make the workaround conditional on firmware version.
This fixes the bug reported by Arthur Kepner <akepner@sgi.com> here:
http://lists.openfabrics.org/pipermail/general/2008-May/050026.html
Cc: <stable@kernel.org>
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
[ Rewritten to be a one-liner using __GFP_ZERO instead of vmap()ing
ICM memory and memset()ing it to 0. - Roland ]
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The mthca driver returns the maximum number of scatter/gather entries
returned by the firmware as the max_sge value when device properties
are queried. However, the firmware also reports a limit on the
maximum descriptor size allowed, and because mthca takes into account
the worst case send request overhead when checking whether to allow a
QP to be created, the largest number of scatter/gather entries that
can be used with mthca may be limited by the maximum descriptor size
rather than just by the actual s/g entry limit.
This means that applications cannot actually create QPs with
max_send_sge equal to the limit returned by ib_query_device(). Fix
this by checking if the maximum descriptor size imposes a lower limit
and if so returning that lower limit.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Commit cb9fbc5c ("IB: expand ib_umem_get() prototype") changed the
mthca userspace ABI to provide a way for userspace to indicate which
memory regions need the DMA write barrier attribute. However, it is
possible to handle this without breaking existing userspace, by having
the mthca kernel driver recognize whether it is talking to old or new
userspace, depending on the size of the register MR structure passed in.
The only potential drawback of this is that is allows old userspace
(which has a bug with DMA ordering on large SGI Altix systems) to
continue to run on new kernels, but the advantage of allowing old
userspace to continue to work on unaffected systems seems to outweigh
this, and we can print a warning to push people to upgrade their
userspace.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
When a FMR is unmapped, mthca resets the map count to 0, and clears
the upper part of the R_Key which is used as the sequence counter.
This poses a problem for RDS, which uses ib_fmr_unmap as a fence
operation. RDS assumes that after issuing an unmap, the old R_Keys
will be invalid for a "reasonable" period of time. For instance,
Oracle processes uses shared memory buffers allocated from a pool of
buffers. When a process dies, we want to reclaim these buffers -- but
we must make sure there are no pending RDMA operations to/from those
buffers. The only way to achieve that is by using unmap and sync the
TPT.
However, when the sequence count is reset on unmap, there is a high
likelihood that a new mapping will be given the same R_Key that was
issued a few milliseconds ago.
To prevent this, don't reset the sequence count when unmapping a FMR.
Signed-off-by: Olaf Kirch <olaf.kirch@oracle.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a new parameter, dmasync, to the ib_umem_get() prototype. Use dmasync = 1
when mapping user-allocated CQs with ib_umem_get().
Signed-off-by: Arthur Kepner <akepner@sgi.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: Jes Sorensen <jes@sgi.com>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: David Miller <davem@davemloft.net>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Grant Grundler <grundler@parisc-linux.org>
Cc: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-2.6: (36 commits)
SCSI: convert struct class_device to struct device
DRM: remove unused dev_class
IB: rename "dev" to "srp_dev" in srp_host structure
IB: convert struct class_device to struct device
memstick: convert struct class_device to struct device
driver core: replace remaining __FUNCTION__ occurrences
sysfs: refill attribute buffer when reading from offset 0
PM: Remove destroy_suspended_device()
Firmware: add iSCSI iBFT Support
PM: Remove legacy PM (fix)
Kobject: Replace list_for_each() with list_for_each_entry().
SYSFS: Explicitly include required header file slab.h.
Driver core: make device_is_registered() work for class devices
PM: Convert wakeup flag accessors to inline functions
PM: Make wakeup flags available whenever CONFIG_PM is set
PM: Fix misuse of wakeup flag accessors in serial core
Driver core: Call device_pm_add() after bus_add_device() in device_add()
PM: Handle device registrations during suspend/resume
block: send disk "change" event for rescan_partitions()
sysdev: detect multiple driver registrations
...
Fixed trivial conflict in include/linux/memory.h due to semaphore header
file change (made irrelevant by the change to mutex).
This converts the main ib_device to use struct device instead of struct
class_device as class_device is going away.
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>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The ib_mthca driver has been stable for a while, so bump the version
number to 1.0 to indicate this.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
If the QP was moved to another state (such as SQE) by the hardware,
then after this change the user won't have to set the IBV_QP_CUR_STATE
mask in order to execute modify QP in order to recover from this state.
Signed-off-by: Dotan Barak <dotanb@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a new IB_WR_SEND_WITH_INV send opcode that can be used to mark a
"send with invalidate" work request as defined in the iWARP verbs and
the InfiniBand base memory management extensions. Also put "imm_data"
and a new "invalidate_rkey" member in a new "ex" union in struct
ib_send_wr. The invalidate_rkey member can be used to pass in an
R_Key/STag to be invalidated. Add this new union to struct
ib_uverbs_send_wr. Add code to copy the invalidate_rkey field in
ib_uverbs_post_send().
Fix up low-level drivers to deal with the change to struct ib_send_wr,
and just remove the imm_data initialization from net/sunrpc/xprtrdma/,
since that code never does any send with immediate operations.
Also, move the existing IB_DEVICE_SEND_W_INV flag to a new bit, since
the iWARP drivers currently in the tree set the bit. The amso1100
driver at least will silently fail to honor the IB_SEND_INVALIDATE bit
if passed in as part of userspace send requests (since it does not
implement kernel bypass work request queueing). Remove the flag from
all existing drivers that set it until we know which ones are OK.
The values chosen for the new flag is not consecutive to avoid clashing
with flags defined in the XRC patches, which are not merged yet but
which are already in use and are likely to be merged soon.
This resurrects a patch sent long ago by Mikkel Hagen <mhagen@iol.unh.edu>.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add a create_flags member to struct ib_qp_init_attr that will allow a
kernel verbs consumer to create a pass special flags when creating a QP.
Add a flag value for telling low-level drivers that a QP will be used
for IPoIB UD LSO. The create_flags member will also be useful for XRC
and ehca low-latency QP support.
Since no create_flags handling is implemented yet, add code to all
low-level drivers to return -EINVAL if create_flags is non-zero.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
In mthca_alloc_icm_table(), the number of entries to allocate for the
table->icm array is computed by calculating obj_size * nobj and then
dividing by MTHCA_TABLE_CHUNK_SIZE. If nobj is really large, then
obj_size * nobj may overflow and the division may get the wrong value
(even a negative value). Fix this by calculating the number of
objects per chunk and then dividing nobj by this value instead.
This patch allows crazy configurations such as loading ib_mthca with
the module parameter num_mtt=33554432 to work properly.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
mthca_make_profile() returns the size in bytes of the HCA context
layout it creates, or a negative value if an error occurs. However,
the return value is declared as u64 and the memfree initialization
path casts this value to int to test if it is negative. This makes it
think incorrectly than an error has occurred if the context size
happens to be bigger than 2GB, since this turns into a negative int.
Fix this by having mthca_make_profile() return an s64 and testing
for an error by checking whether this 64-bit value itself is negative.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Arbel and Sinai devices support checksum generation and verification
of TCP and UDP packets for UD IPoIB messages. This patch checks if
the HCA supports this and sets the IB_DEVICE_UD_IP_CSUM capability
flag if it does. It implements support for handling the IB_SEND_IP_CSUM
send flag and setting the csum_ok field in receive work completions.
Signed-off-by: Eli Cohen <eli@mellnaox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
When mthca_fmr_alloc() returns an error, it should free the MPT at the
index key, not mr->ibmr.lkey, since the lkey has been mangled by
hw_index_to_key() and no longer is the real index. This bug causes
corruption of the MPT table free bitmap when mthca_fmr_alloc() fails.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
replace:
big_endian_variable = cpu_to_beX(beX_to_cpu(big_endian_variable) +
expression_in_cpu_byteorder);
with:
beX_add_cpu(&big_endian_variable, expression_in_cpu_byteorder);
Generated with a semantic patch.
Signed-off-by: Marcin Slusarz <marcin.slusarz@gmail.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Hal Rosenstock <hal.rosenstock@gmail.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Usually harmless, since the scatterlist is always hard-coded to a length
of 1, but it triggers a BUG() if CONFIG_DEBUG_SG=y, so we better fix it.
This fixes <http://bugzilla.kernel.org/show_bug.cgi?id=9934>.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
If the allocation of the MTT or the mailbox failed, mthca_fmr_alloc()
would return 0 (success) no matter what. This leads to crashes a
little down the road, when we try to dereference eg mr->mtt, which was
really ERR_PTR(-Ewhatever).
Signed-off-by: Olaf Kirch <olaf.kirch@oracle.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
We have recently discovered that Tavor mode requires each WQE in a
posted list of receive WQEs to have a valid NDA field at all times.
This requirement holds true for regular QPs as well as for SRQs. This
patch prelinks the receive queue in a regular QP and keeps the free
list in SRQ always properly linked.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Reviewed-by: Jack Morgenstein <jackm@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The SRQ receive posting functions make sure that srq->first_free never
becomes negative, so we can remove tests of whether it is negative.
Signed-off-by: Eli Cohen <eli@mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
For memfree devices, the firmware QUERY_ADAPTER command does not
return vendor_id, device_id, and revision_id; do not return these
fields in the QUERY_ADAPTER function for memfree devices.
Instead, for memfree devices, initialize the rev_id field of the mthca
device via init_node_data (MAD IFC query), as is done in the
query_device verb implementation.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
In mthca_reg_phys_mr(), we calculate the page size for the HCA
hardware to use to map the buffer list passed in by the consumer.
For example, if the consumer passes in
[0] addr 0x1000, size 0x1000
[1] addr 0x2000, size 0x1000
then the algorithm would come up with a page size of 0x2000 and a list
of two pages, at 0x0000 and 0x2000. Usually, this would work fine
since the memory region would start at an offset of 0x1000 and have a
length of 0x2000.
However, the old code did not take into account the alignment of the
IO virtual address passed in. For example, if the consumer passed in
a virtual address of 0x6000 for the above, then the offset of 0x1000
would not be used correctly because the page mask of 0x1fff would
result in an offset of 0.
We can fix this quite neatly by making sure that the page shift we use
is no bigger than the first bit where the start of the first buffer
and the IO virtual address differ. Also, we can further simplify the
code by removing the special case for a single buffer by noticing that
it doesn't matter if we use a page size that is too big. This allows
the loop to compute the page shift to be replaced with __ffs().
Thanks to Bryan S Rosenburg <rosnbrg@us.ibm.com> for pointing out the
original bug and suggesting several ways to improve this patch.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Remove MSI support from the mthca driver, as scheduled. There is no
reason to use MSI instead of MSI-X, since MSI-X performs better. No
one has spoken up since MSI support was deprecated in commit f6be6fbe
("IB/mthca: Schedule MSI support for removal"), so apparently the MSI
support is unused.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Most drivers need to set length and offset as well, so may as well fold
those three lines into one.
Add sg_assign_page() for those two locations that only needed to set
the page, where the offset/length is set outside of the function context.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
mlx4_core: Increase command timeout for INIT_HCA to 10 seconds
IPoIB/cm: Use common CQ for CM send completions
IB/uverbs: Fix checking of userspace object ownership
IB/mlx4: Sanity check userspace send queue sizes
IPoIB: Rewrite "if (!likely(...))" as "if (unlikely(!(...)))"
IB/ehca: Enable large page MRs by default
IB/ehca: Change meaning of hca_cap_mr_pgsize
IB/ehca: Fix ehca_encode_hwpage_size() and alloc_fmr()
IB/ehca: Fix masking error in {,re}reg_phys_mr()
IB/ehca: Supply QP token for SRQ base QPs
IPoIB: Use round_jiffies() for ah_reap_task
RDMA/cma: Fix deadlock destroying listen requests
RDMA/cma: Add locking around QP accesses
IB/mthca: Avoid alignment traps when writing doorbells
mlx4_core: Kill mlx4_write64_raw()
Architectures such as ia64 see alignment traps when doing a 64-bit
read from __be32 doorbell[2] arrays to do doorbell writes in
mthca_write64(). Fix this by just passing the two halves of the
doorbell value into mthca_write64(). This actually improves the
generated code by allowing the compiler to see what's going on better.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Firmware commands are sent to the HCA by writing multiple words to a
command register block. Access to this block of registers is
serialized with a mutex. However, on large SGI systems, problems were
seen with multiple CPUs issuing FW commands at the same time, because
the writes to the register block may be reordered within the system
interconnect and reach the HCA in a different order than they were
issued (even with the mutex). Fix this by adding an mmiowb() before
dropping the mutex.
Tested-by: Arthur Kepner <akepner@sgi.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Increase the number of QPs allowed per multicast group from 8 to 56.
This allows for one QP per core on 16-core systems, which are now
quite common, and allows some space for future growth.
This is basically the same patch that Jack Morgenstein
<jackm@dev.mellanox.co.il> just supplied for mlx4.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
These driver changes incorporate the proposed PCI-X / PCI-Express read
byte count interface. Reading and setting those values doesn't take
place "manually", instead wrapping functions are called to allow
quirks for some PCI bridges.
Signed-off by: Peter Oruba <peter.oruba@amd.com>
Based on work by Stephen Hemminger <shemminger@linux-foundation.org>
Cc: Roland Dreier <rolandd@cisco.com>
Cc: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Recover from MSI-X errors by automatically falling back on regular
interrupt, instead of asking the user to do this manually. This makes
it possible to enable MSI-X by default, and will make it possible to
get rid of the msi_x module option in the future.
Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The FW command token is currently only updated on a command completion
event. This means that on command timeout, the same token will be
reused for new command, which results in a mess if the timed out
command *does* eventually complete.
Signed-off-by: Michael S. Tsirkin <mst@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Current code sets size0 to 0 at the start of work request posting
functions and then handles size0 == 0 specially within the loop over
work requests. Change this so size0 is set along with f0 the first
time through the loop (when nreq == 0). This makes the code easier to
understand by making it clearer that f0 and size0 are always
initialized if nreq != 0 without having to know that size0 == 0
implies nreq == 0.
Also annotate size0 with uninitialized_var() so that this doesn't
introduce a new compiler warning.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Factor code to set UD entries out of the work request posting
functions into inline functions set_tavor_ud_seg() and
set_arbel_ud_seg(). This doesn't change the generated code in any
significant way, and makes the source easier on the eyes.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Factor code to set remote address and atomic segment entries out of the
work request posting functions into inline functions set_raddr_seg()
and set_atomic_seg(). This doesn't change the generated code in any
significant way, and makes the source easier on the eyes.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Factor code to set data segment entries out of the work request
posting functions into inline functions mthca_set_data_seg() and
mthca_set_data_seg_inval(). This makes the code more readable and
also allows the compiler to do a better job -- on x86_64:
add/remove: 0/0 grow/shrink: 0/6 up/down: 0/-69 (-69)
function old new delta
mthca_arbel_post_srq_recv 373 369 -4
mthca_arbel_post_receive 570 562 -8
mthca_tavor_post_srq_recv 520 508 -12
mthca_tavor_post_send 1344 1330 -14
mthca_arbel_post_send 1481 1467 -14
mthca_tavor_post_receive 792 775 -17
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Commit 9db48926 ("drivers/infiniband/hw/mthca/mthca_qp: kill uninit'd
var warning") added "= 0" to the declarations of f0 to shut up gcc
warnings. However, there's no point in making the code bigger by
initializing f0 to a random value just to get rid of a warning;
setting f0 to 0 is no safer than just using uninitialized_var(), which
documents the situation better and gives smaller code too. For example,
on x86_64:
add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-16 (-16)
function old new delta
mthca_tavor_post_send 1352 1344 -8
mthca_arbel_post_send 1489 1481 -8
Signed-off-by: Roland Dreier <rolandd@cisco.com>
When warning about out-of-date firmware, current mthca code messes up
the formatting of the version if the subminor doesn't have three
digits. It doesn't fill the field with 0s so we end up with:
ib_mthca 0000:0b:00.0: HCA FW version 1.1. 0 is old (1.2. 0 is current).
Change the format from "%3d" to "%03d" to get the right thing printed.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The mthca driver supports both MSI and MSI-X. However, MSI-X works with
all hardware that the driver handles, and provides a superset of what
MSI does, so there's no point in having code for both. Schedule MSI
support for removal in 2008 to give anyone who actually needs MSI and
who can't use MSI time to speak up.
Signed-off-by: Roland Dreier <rolandd@cisco.com>
