Pierre Ossman [Sun, 29 Jun 2008 10:19:47 +0000 (12:19 +0200)]
mmc_block: wait for card even on failures
Many failures are non-permanent, but the card might need some time to
finish what it is doing before becoming responsive again. Make sure we
wait for it to finish programming before dealing with the error.
Pierre Ossman [Sat, 28 Jun 2008 16:28:51 +0000 (18:28 +0200)]
sdhci: scatter-gather (ADMA) support
Add support for the scatter-gather DMA mode present on newer controllers.
As the mode requires 32-bit alignment, non-aligned chunks are handled by
using a bounce buffer.
Also add some new quirks to handle controllers that have bugs in the
ADMA engine.
Pierre Ossman [Sat, 28 Jun 2008 11:22:40 +0000 (13:22 +0200)]
sdio: clean up handling of byte mode transfer size
Make sure that the maximum size for a byte mode transfer is identical
in all places. Also tweak the transfer helper so that a single byte
mode transfer is preferred over (possibly multiple) block mode
request(s).
Pierre Ossman [Sat, 28 Jun 2008 10:52:45 +0000 (12:52 +0200)]
mmc,sdio: helper function for transfer padding
There are a lot of crappy controllers out there that cannot handle
all the request sizes that the MMC/SD/SDIO specifications require.
In case the card driver can pad the data to overcome the problems,
this commit adds a helper that calculates how much that padding
should be.
A corresponding helper is also added for SDIO, but it can also deal
with all the complexities of splitting up a large transfer efficiently.
Manuel Lauss [Mon, 9 Jun 2008 06:36:13 +0000 (08:36 +0200)]
au1xmmc: remove db1200 board code, rewrite probe.
Remove the DB1200 board-specific functions (card present, read-only,
activity LED methods) and instead add platform data which is passed
to the driver. This also allows for platforms to implement other
carddetect schemes (e.g. dedicated irq) without having to pollute the
driver code. The poll timer (used for pb1200) is kept for compatibility.
With the board-specific stuff gone, the driver's ->probe() code can be
cleaned up considerably.
Signed-off-by: Manuel Lauss <mano@roarinelk.homelinux.net> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Ville Syrjala [Mon, 9 Jun 2008 19:06:45 +0000 (22:06 +0300)]
at91_mci: Fix byte mode transitions.
The byte mode support fails to clear the byte mode bit in the command
register, possibly leaving byte mode enabled with the counters programmed
in non-byte mode.
Signed-off-by: Ville Syrjala <syrjala@sci.fi> Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Ville Syrjala [Mon, 9 Jun 2008 19:06:44 +0000 (22:06 +0300)]
at91_mci: Cover more AT91RM9200 and AT91SAM9261 errata.
According to the documentation the AT91SAM9261 MCI shares the block size
limitations of the AT91RM9200 MCI. Also the errata documentation for
AT91RM9200 and AT91SAM9261 state that stream commands are not supported.
This has not been tested on actual hardware.
Signed-off-by: Ville Syrjala <syrjala@sci.fi> Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
AT91SAM926[0/3] PDC must write at least 12 bytes. The code compiles and runs
but the actual condition for this erratum did not trigger in my tests so it's
unclear if it actually works as intended.
Signed-off-by: Ville Syrjala <syrjala@sci.fi> Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Marc Pignat [Fri, 30 May 2008 12:07:47 +0000 (14:07 +0200)]
mmc: at91_mci: avoid timeouts
The at91 mci controller internal state machine seems to often crash. This can
be fixed by resetting the controller after each command for at91rm9200 and by
setting the MCI_BLKR register on at91sam926*.
Signed-off-by: Marc Pignat <marc.pignat@hevs.ch> Signed-off-by: Hans J Koch <hjk@linutronix.de> Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Marc Pignat [Fri, 30 May 2008 12:06:32 +0000 (14:06 +0200)]
mmc: at91_mci: show timeouts
Detect command timeout (or mci controller hangs).
Signed-off-by: Marc Pignat <marc.pignat@hevs.ch> Signed-off-by: Hans J Koch <hjk@linutronix.de> Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Marc Pignat [Fri, 30 May 2008 12:05:24 +0000 (14:05 +0200)]
mmc: at91_mci: support for block size not modulo 4
Implement transfer with size not modulo 4 for at91sam9*. Please note that the
at91rm9200 simply can't handle this.
Signed-off-by: Marc Pignat <marc.pignat@hevs.ch> Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Anton Vorontsov [Tue, 17 Jun 2008 14:17:39 +0000 (18:17 +0400)]
mmc: change .get_ro() callback semantics
Now get_ro() callback must return 0/1 values for its logical states, and
negative errno values in case of error. If particular host instance doesn't
support RO/WP switch, it should return -ENOSYS.
This patch changes some hosts in two ways:
1. Now functions should be smart to not return negative values in
"RO asserted" case (particularly gpio_ calls could return negative
values for the outermost GPIOs).
Also, board code usually passes get_ro() callbacks that directly return
gpioreg & bit result, so at91_mci, imxmmc, pxamci and mmc_spi's get_ro()
handlers need take special care when returning platform's values to the
mmc core.
2. In case of host instance didn't implement get_ro() callback, it should
really return -ENOSYS and let the mmc core decide what to do about it
(mmc core thinks the same way as the hosts, so it isn't functional
change).
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Anton Vorontsov [Tue, 17 Jun 2008 14:17:21 +0000 (18:17 +0400)]
mmc_spi: add support for card-detection polling
This patch adds new platform data variable "caps", so platforms
could pass theirs capabilities into MMC core (for example, platforms
without interrupt on the CD line will most probably want to pass
MMC_CAP_NEEDS_POLL).
New platform get_cd() callback provided to optimize polling.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Anton Vorontsov [Tue, 17 Jun 2008 14:17:15 +0000 (18:17 +0400)]
mmc: add support for card-detection polling
Some hosts (and boards that use mmc_spi) do not use interrupts on the CD
line, so they can't trigger mmc_detect_change. We want to poll the card
and see if there was a change. 1 second poll interval seems resonable.
This patch also implements .get_cd() host operation, that could be used
by the hosts that are able to report card-detect status without need to
talk MMC.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com> Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
Pierre Ossman [Fri, 4 Jul 2008 22:25:15 +0000 (00:25 +0200)]
sdhci: make workaround for timeout bug more general
Give the quirk for broken timeout handling a better chance of handling
more controllers by simply classifying the system as broken and setting
a fixed value.
Pierre Ossman [Sat, 22 Mar 2008 21:05:40 +0000 (22:05 +0100)]
sdhci: remove forced dma quirks
Remove the quirk to force DMA on the Ricoh and TI controllers as it is
no longer needed. The only bug they have is that they use an incorrect
PCI interface value, and that is not respected anymore.
Pierre Ossman [Tue, 18 Mar 2008 16:35:49 +0000 (17:35 +0100)]
sdhci: move pci stuff to separate module
The SDHCI interface is not PCI specific, yet the Linux driver was
intimitely connected to the PCI bus. This patch properly separates
the PCI specific portion from the bus independent code.
This patch is based on work by Ben Dooks but he did not have time
to complete it.
Pierre Ossman [Fri, 18 Apr 2008 18:41:49 +0000 (20:41 +0200)]
sdhci: don't check block count for progress
The specification is insufficiently strict when it comes to how the
hardware should update the block count register, making it useless
for checking transfer progress.
Peter Zijlstra [Tue, 22 Apr 2008 13:09:30 +0000 (15:09 +0200)]
Kprobe smoke test lockdep warning
On Mon, 2008-04-21 at 18:54 -0400, Masami Hiramatsu wrote:
> Thank you for reporting.
>
> Actually, kprobes tries to fixup thread's flags in post_kprobe_handler
> (which is called from kprobe_exceptions_notify) by
> trace_hardirqs_fixup_flags(pt_regs->flags). However, even the irq flag
> is set in pt_regs->flags, true hardirq is still off until returning
> from do_debug. Thus, lockdep assumes that hardirq is off without annotation.
>
> IMHO, one possible solution is that fixing hardirq flags right after
> notify_die in do_debug instead of in post_kprobe_handler.
Ok, do_debug() is a trap, this can happen at any time regardless of the
machine's IRQ state. So the first thing we do is fix up the IRQ state.
Then we call this die notifier stuff; and return with messed up IRQ
state... YAY.
So, kprobes fudges it..
notify_die(DIE_DEBUG)
kprobe_exceptions_notify()
post_kprobe_handler()
modify regs->flags
trace_hardirqs_fixup_flags(regs->flags); <--- must be it
So what's the use of modifying flags if they're not meant to take effect
at some point.
/me tries to reproduce issue; enable kprobes test thingy && boot
OK, that reproduces..
So the below makes it work - but I'm not getting this code; at the time
I wrote that stuff I CC'ed each and every kprobe maintainer listed in
the usual places but got no reposonse - can some please explain this
stuff to me?
Are the saved flags only for the TF bit or are they made in full effect
later (and if so, where) ?
Johannes Berg [Tue, 15 Jul 2008 09:08:24 +0000 (02:08 -0700)]
wireless: fix warnings from QoS patch
When I removed the special "default" meaning from the QoS
parameters, I forgot to update drivers and this lead to
warnings because some drivers were checking for the special
values and putting in defaults. This fixes that by removing
the default special-casing completely.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Randy Dunlap [Tue, 15 Jul 2008 07:51:45 +0000 (00:51 -0700)]
bluetooth/hci_bcsp: fix bitrev Kconfig
Fix bluetooth hci_bcsp Kconfig to avoid build errors:
drivers/built-in.o: In function `bcsp_prepare_pkt':
hci_bcsp.c:(.text+0x7e9ac): undefined reference to `bitrev16'
drivers/built-in.o: In function `bcsp_recv':
hci_bcsp.c:(.text+0x7f276): undefined reference to `bitrev16'
hci_bcsp.c:(.text+0x7f293): undefined reference to `bitrev16'
make[1]: *** [.tmp_vmlinux1] Error 1
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com> Ackey-by: Marcel Holtmann <marcel@holtmann.org> Signed-off-by: David S. Miller <davem@davemloft.net>
net: refactor tcp splice receive path to improve readability
- move all of the details on offsets, lengths and buffers into a
single function instead of doing these operation from multiple places
- use a bottom up approach: try to avoid details in the high level
functions, introduce them gradually as we go deeper in the function
call stack
With helpful feedback from Jarek Poplawski.
Signed-off-by: Octavian Purdila <opurdila@ixiacom.com> Acked-by: Jarek Poplawski <jarkao2@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Tue, 15 Jul 2008 07:15:08 +0000 (00:15 -0700)]
netdev: Do not use TX lock to protect address lists.
Now that we have a specific lock to protect the network
device unicast and multicast lists, remove extraneous
grabs of the TX lock in cases where the code only needs
address list protection.
Signed-off-by: David S. Miller <davem@davemloft.net>
David S. Miller [Tue, 15 Jul 2008 07:08:33 +0000 (00:08 -0700)]
netdev: Add addr_list_lock to struct net_device.
This will be used to protect the per-device unicast and multicast
address lists, as well as the callbacks into the drivers which
configure such state such as ->set_rx_mode() and ->set_multicast_list().
Signed-off-by: David S. Miller <davem@davemloft.net>
Eli Cohen [Tue, 15 Jul 2008 06:48:53 +0000 (23:48 -0700)]
IB/mlx4: Use kzalloc() for new QPs so flags are initialized to 0
Current code uses kmalloc() and then just does a bitwise OR operation on
qp->flags in create_qp_common(), which means that qp->flags may
potentially have some unintended bits set. This patch uses kzalloc()
and avoids further explicit clearing of structure members, which also
shrinks the code:
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-65 (-65)
function old new delta
create_qp_common 2024 1959 -65
Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
mlx4_core: Use MOD_STAT_CFG command to get minimal page size
There was a bug in some versions of the mlx4 driver in
mlx4_alloc_fmr(), which hardcoded the minimum acceptable page_shift to
be 12. However, new ConnectX firmware can support a minimum
page_shift of 9 (log_pg_sz of 9 returned by QUERY_DEV_LIM) -- so with
old drivers, ib_fmr_alloc() would fail for ULPs using the device
minimum when creating FMRs.
To preserve firmware compatibility with released mlx4 drivers, the
firmware will continue to return 12 as before for log_page_sz in
QUERY_DEV_CAP for these drivers. However, to enable new drivers to
take advantage of the available smaller page size, the mlx4 driver now
first sets the log_pg_sz to the device minimum by setting a
log_page_sz value to 0 via the MOD_STAT_CFG command and then reading
the real minimum via QUERY_DEV_CAP.
Signed-off-by: Jack Morgenstein <jackm@mellanox.co.il> Signed-off-by: Vladimir Sokolovsky <vlad@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Or Gerlitz [Tue, 15 Jul 2008 06:48:53 +0000 (23:48 -0700)]
RDMA/cma: Simplify locking needed for serialization of callbacks
The RDMA CM has some logic in place to make sure that callbacks on a
given CM ID are delivered to the consumer in a serialized manner.
Specifically it has code to protect against a device removal racing
with a running callback function.
This patch simplifies this logic by using a mutex per ID instead of a
wait queue and atomic variable. This means that cma_disable_remove()
now is more properly named to cma_disable_callback(), and
cma_enable_remove() can now be removed because it just would become a
trivial wrapper around mutex_unlock().
Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Or Gerlitz [Tue, 15 Jul 2008 06:48:53 +0000 (23:48 -0700)]
RDMA/addr: Keep pointer to netdevice in struct rdma_dev_addr
Keep a pointer to the local (src) netdevice in struct rdma_dev_addr,
and copy it in as part of rdma_copy_addr(). Use rdma_translate_ip()
in cma_new_conn_id() to reduce some code duplication and also make
sure the src_dev member gets set.
In a high-availability configuration the netdevice pointer can be used
by the RDMA CM to align RDMA sessions to use the same links as the IP
stack does under fail-over and route change cases.
Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Steve Wise [Tue, 15 Jul 2008 06:48:53 +0000 (23:48 -0700)]
RDMA/cxgb3: Fixes for zero STag
Handling the zero STag in receive work request requires some extra
logic in the driver:
- Only set the QP_PRIV bit for kernel mode QPs.
- Add a zero STag build function for recv wrs. The uP needs a PBL
allocated and passed down in the recv WR so it can construct a HW
PBL for the zero STag S/G entries. Note: we need to place a few
restrictions on zero STag usage because of this:
1) all SGEs in a recv WR must either be zero STag or not. No mixing.
2) an individual SGE length cannot exceed 128MB for a zero-stag SGE.
This should be OK since it's not really practical to allocate
such a large chunk of pinned contiguous DMA mapped memory.
- Add an optimized non-zero-STag recv wr format for kernel users.
This is needed to optimize both zero and non-zero STag cracking in
the recv path for kernel users.
- Remove the iwch_ prefix from the static build functions.
- Bump required FW version.
Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Steve Wise [Tue, 15 Jul 2008 06:48:53 +0000 (23:48 -0700)]
RDMA/core: Add local DMA L_Key support
- Change the IB_DEVICE_ZERO_STAG flag to the transport-neutral name
IB_DEVICE_LOCAL_DMA_LKEY, which is used by iWARP RNICs to indicate 0
STag support and IB HCAs to indicate reserved L_Key support.
- Add a u32 local_dma_lkey member to struct ib_device. Drivers fill
this in with the appropriate local DMA L_Key (if they support it).
- Fix up the drivers using this flag.
Signed-off-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Roland Dreier [Tue, 15 Jul 2008 06:48:52 +0000 (23:48 -0700)]
IB/mthca: Fix check of max_send_sge for special QPs
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.
Eli Cohen [Tue, 15 Jul 2008 06:48:52 +0000 (23:48 -0700)]
IPoIB: Double default RX/TX ring sizes
Increase IPoIB ring sizes to twice their original sizes (RX: 128->256,
TX: 64->128) to act as a shock absorber for high traffic peaks. With
the current settings, we have seen cases that there are many calls to
netif_stop_queue(), which causes degradation in throughput. Also,
larger receive buffer sizes help IPoIB in CM mode to avoid experiencing
RNR NAK conditions due to insufficient receive buffers at the SRQ.
Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Eli Cohen [Tue, 15 Jul 2008 06:48:52 +0000 (23:48 -0700)]
IPoIB/cm: Reduce connected mode TX object size
Since IPoIB connected mode does not NETIF_F_SG, we only have one DMA
mapping per send, so we don't need a mapping[] array. Define a new
struct with a single u64 mapping member and use it for the CM tx_ring.
Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Ralph Campbell [Tue, 15 Jul 2008 06:48:52 +0000 (23:48 -0700)]
IB/ipath: Use IEEE OUI for vendor_id reported by ibv_query_device()
The IB spe. for SubnGet(NodeInfo) and query HCA says that the vendor
ID field should be the IEEE OUI assigned to the vendor. The ipath
driver was returning the PCI vendor ID instead. This will affect
applications which call ibv_query_device(). The old value was
0x001fc1 or 0x001077, the new value is 0x001175.
The vendor ID doesn't appear to be exported via /sys so that should
reduce possible compatibility issues. I'm only aware of Open MPI as a
major application which depends on this change, and they have made
necessary adjustments.
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Eli Cohen [Tue, 15 Jul 2008 06:48:51 +0000 (23:48 -0700)]
IPoIB: Use dev_set_mtu() to change mtu
When the driver sets the MTU of the net device outside of its
change_mtu method, it should make use of dev_set_mtu() instead of
directly setting the mtu field of struct netdevice. Otherwise
functions registered to be called upon MTU change will not get called
(this is done through call_netdevice_notifiers() in dev_set_mtu()).
Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Eli Cohen [Tue, 15 Jul 2008 06:48:51 +0000 (23:48 -0700)]
IPoIB: Use rtnl lock/unlock when changing device flags
Use of this lock is required to synchronize changes to the netdvice's
data structs. Also move the call to ipoib_flush_paths() after the
modification of the netdevice flags in set_mode().
Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Eli Cohen [Tue, 15 Jul 2008 06:48:50 +0000 (23:48 -0700)]
IPoIB: Only set Q_Key once: after joining broadcast group
The current code will set the Q_Key for any join of a non-sendonly
multicast group. The operation involves a modify QP operation, which
is fairly heavyweight, and is only really required after the join of
the broadcast group. Fix this by adding a parameter to ipoib_mcast_attach()
to control when the Q_Key is set.
Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Jon Mason [Tue, 15 Jul 2008 06:48:49 +0000 (23:48 -0700)]
RDMA/cxgb3: Propagate HW page size capabilities
cxgb3 does not currently report the page size capabilities, and
incorrectly reports them internally.
This version changes the bit-shifting to a static value (per Steve's
request).
Signed-off-by: Jon Mason <jon@opengridcomputing.com> Acked-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Moni Shoua [Tue, 15 Jul 2008 06:48:49 +0000 (23:48 -0700)]
IPoIB: Refresh paths instead of flushing them on SM change events
The patch tries to solve the problem of device going down and paths being
flushed on an SM change event. The method is to mark the paths as candidates for
refresh (by setting the new valid flag to 0), and wait for an ARP
probe a new path record query.
The solution requires a different and less intrusive handling of SM
change event. For that, the second argument of the flush function
changes its meaning from a boolean flag to a level. In most cases, SM
failover doesn't cause LID change so traffic won't stop. In the rare
cases of LID change, the remote host (the one that hadn't changed its
LID) will lose connectivity until paths are refreshed. This is no
worse than the current state. In fact, preventing the device from
going down saves packets that otherwise would be lost.
Signed-off-by: Moni Levy <monil@voltaire.com> Signed-off-by: Moni Shoua <monis@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Add "ipoib_use_lro" module parameter to enable LRO and an
"ipoib_lro_max_aggr" module parameter to set the max number of packets
to be aggregated. Make LRO controllable and LRO statistics accessible
through ethtool.
Signed-off-by: Vladimir Sokolovsky <vlad@mellanox.co.il> Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Ron Livne [Tue, 15 Jul 2008 06:48:48 +0000 (23:48 -0700)]
IPoIB: Use multicast loopback blocking if available
Set IB_QP_CREATE_BLOCK_MULTICAST_LOOPBACK for IPoIB's UD QPs if
supported by the underlying device. This creates an improvement of up
to 39% in bandwidth when sending multicast packets with IPoIB, and an
improvment of 12% in cpu usage.
Signed-off-by: Ron Livne <ronli@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Ron Livne [Tue, 15 Jul 2008 06:48:48 +0000 (23:48 -0700)]
IB/core: Add support for multicast loopback blocking
This patch also adds a creation flag for QPs,
IB_QP_CREATE_MULTICAST_BLOCK_LOOPBACK, which when set means that
multicast sends from the QP to a group that the QP is attached to will
not be looped back to the QP's receive queue. This can be used to
save receive resources when a consumer does not want a local copy of
multicast traffic; for example IPoIB must waste CPU time throwing away
such local copies of multicast traffic.
This patch also adds a device capability flag that shows whether a
device supports this feature or not.
Signed-off-by: Ron Livne <ronli@voltaire.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Steve Wise [Tue, 15 Jul 2008 06:48:48 +0000 (23:48 -0700)]
RDMA/core: Add iWARP protocol statistics attributes in sysfs
This patch adds a sysfs attribute group called "proto_stats" under
/sys/class/infiniband/$device/ and populates this group with protocol
statistics if they exist for a given device. Currently, only iWARP
stats are defined, but the code is designed to allow InfiniBand
protocol stats if they become available. These stats are per-device
and more importantly -not- per port.
Details:
- Add union rdma_protocol_stats in ib_verbs.h. This union allows
defining transport-specific stats. Currently only iwarp stats are
defined.
- Add struct iw_protocol_stats to define the current set of iwarp
protocol stats.
- Add new ib_device method called get_proto_stats() to return protocol
statistics.
- Add logic in core/sysfs.c to create iwarp protocol stats attributes
if the device is an RNIC and has a get_proto_stats() method.
Signed-off-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Roland Dreier [Tue, 15 Jul 2008 06:48:47 +0000 (23:48 -0700)]
IPoIB/cm: Fix racy use of receive WR/SGL in ipoib_cm_post_receive_nonsrq()
For devices that don't support SRQs, ipoib_cm_post_receive_nonsrq() is
called from both ipoib_cm_handle_rx_wc() and ipoib_cm_nonsrq_init_rx(),
and these two callers are not synchronized against each other.
However, ipoib_cm_post_receive_nonsrq() always reuses the same receive
work request and scatter list structures, so multiple callers can end
up stepping on each other, which leads to posting garbled work
requests.
Fix this by having the caller pass in the ib_recv_wr and ib_sge
structures to use, and allocating new local structures in
ipoib_cm_nonsrq_init_rx().
Based on a patch by Pradeep Satyanarayana <pradeep@us.ibm.com> and
David Wilder <dwilder@us.ibm.com>, with debugging help from Hoang-Nam
Nguyen <hnguyen@de.ibm.com>.
Stefan Roscher [Tue, 15 Jul 2008 06:48:47 +0000 (23:48 -0700)]
IB/ehca: In case of lost interrupts, trigger EOI to reenable interrupts
During corner case testing, we noticed that some versions of ehca do
not properly transition to interrupt done in special load situations.
This can be resolved by periodically triggering EOI through H_EOI, if
EQEs are pending.
Signed-off-by: Stefan Roscher <stefan.roscher@de.ibm.com> Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Roland Dreier [Tue, 15 Jul 2008 06:48:46 +0000 (23:48 -0700)]
IB/mlx4: Remove extra code for RESET->ERR QP state transition
Commit 65adfa91 ("IB/mlx4: Fix RESET to RESET and RESET to ERROR
transitions") 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.
Roland Dreier [Tue, 15 Jul 2008 06:48:46 +0000 (23:48 -0700)]
IB/mthca: Remove extra code for RESET->ERR QP state transition
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.
Ralph Campbell [Tue, 15 Jul 2008 06:48:46 +0000 (23:48 -0700)]
IB/core: Reset to error QP state transition is not allowed
I was reviewing the QP state transition diagram in the IB 1.2.1 spec
and the code for qp_state_table[], and noticed that the code allows a
QP to be modified from IB_QPS_RESET to IB_QPS_ERR whereas the notes
for figure 124 (pg 457) specifically says that this transition isn't
allowed. This is a clarification from earlier versions of the IB
spec, which were ambiguous in this area and suggested that the RESET
to ERR transition was allowed.
Fix up the qp_state_table[] to make RESET->ERR not allowed.
Signed-off-by: Ralph Campbell <ralph.campbell@qlogic.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Eli Cohen [Tue, 15 Jul 2008 06:48:45 +0000 (23:48 -0700)]
IB/mlx4: Configure QPs' max message size based on real device capability
ConnectX returns the max message size it supports through the
QUERY_DEV_CAP firmware command. When modifying a QP to RTR, the max
message size for the QP must be specified. This value must not exceed
the value declared through QUERY_DEV_CAP. The current code ignores
the max allowed size and unconditionally sets the value to 2^31. This
patch sets all QPs to the max value allowed as returned from firmware.
Signed-off-by: Eli Cohen <eli@mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Steve Wise [Tue, 15 Jul 2008 06:48:45 +0000 (23:48 -0700)]
RDMA/cxgb3: MEM_MGT_EXTENSIONS support
- set IB_DEVICE_MEM_MGT_EXTENSIONS capability bit if fw supports it.
- set max_fast_reg_page_list_len device attribute.
- add iwch_alloc_fast_reg_mr function.
- add iwch_alloc_fastreg_pbl
- add iwch_free_fastreg_pbl
- adjust the WQ depth for kernel mode work queues to account for
fastreg possibly taking 2 WR slots.
- add fastreg_mr work request support.
- add local_inv work request support.
- add send_with_inv and send_with_se_inv work request support.
- removed useless duplicate enums/defines for TPT/MW/MR stuff.
Signed-off-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <rolandd@cisco.com>
Steve Wise [Tue, 15 Jul 2008 06:48:45 +0000 (23:48 -0700)]
RDMA/core: Add memory management extensions support
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>
Eli Cohen [Tue, 15 Jul 2008 06:48:44 +0000 (23:48 -0700)]
IPoIB: Copy small received SKBs in connected mode
The connected mode implementation in the IPoIB driver has a large
overhead in the way SKBs are handled in the receive flow. It usually
allocates an SKB with as big as was used in the currently received SKB
and moves unused fragments from the old SKB to the new one. This
involves a loop on all the remaining fragments and incurs overhead on
the CPU. This patch, for small SKBs, allocates an SKB just large
enough to contain the received data and copies to it the data from the
received SKB. The newly allocated SKB is passed to the stack and the
old SKB is reposted.
When running netperf, UDP small messages, without this pach I get:
UDP UNIDIRECTIONAL SEND TEST from 0.0.0.0 (0.0.0.0) port 0 AF_INET to
14.4.3.178 (14.4.3.178) port 0 AF_INET
Socket Message Elapsed Messages
Size Size Time Okay Errors Throughput
bytes bytes secs # # 10^6bits/sec
With this patch I get both send and receive at ~315 mbps.
The reason that send performance actually slows down is as follows:
When using this patch, the overhead of the CPU for handling RX packets
is dramatically reduced. As a result, we do not experience RNR NAK
messages from the receiver which cause the connection to be closed and
reopened again; when the patch is not used, the receiver cannot handle
the packets fast enough so there is less time to post new buffers and
hence the mentioned RNR NACKs. So what happens is that the
application *thinks* it posted a certain number of packets for
transmission but these packets are flushed and do not really get
transmitted. Since the connection gets opened and closed many times,
each time netperf gets the CPU time that otherwise would have been
given to IPoIB to actually transmit the packets. This can be verified
when looking at the port counters -- the output of ifconfig and the
oputput of netperf (this is for the case without the patch):
tx packets
==========
port counter: 1,543,996
ifconfig: 1,581,426
netperf: 5,142,034
projects / linux-2.6-omap-h63xx.git / log