Mike Marciniszyn [Sun, 14 Feb 2016 20:45:53 +0000 (12:45 -0800)]
staging/rdma/hfi: fix CQ completion order issue
The current implementation of the sdma_wait variable
has a timing hole that can cause a completion Q entry
to be returned from a pio send prior to an older
sdma packets completion queue entry.
The sdma_wait variable used to be decremented prior to
calling the packet complete routine. The hole is between decrement
and the verbs completion where send engine using pio could return
a out of order completion in that window.
This patch closes the hole by allowing an API option to
specify an sdma_drained callback. The atomic dec
is positioned after the complete callback to avoid the
window as long as the pio path doesn't execute when
there is a non-zero sdma count.
Reviewed-by: Jubin John <jubin.john@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
The non-rdamvt versions of qib and hfi1 allow for a differing
heuristic to override a schedule progress in favor of a direct
call the the progress routine.
This patch adds that for both drivers and rdmavt.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mike Marciniszyn [Sun, 14 Feb 2016 20:45:36 +0000 (12:45 -0800)]
staging/rdma/hfi1: Adaptive PIO for short messages
The change requires a new pio_busy field in the iowait structure to
track the number of outstanding pios. The new counter together
with the sdma counter serve as the basis for a packet by packet decision
as to which egress mechanism to use. Since packets given to different
egress mechanisms are not ordered, this scheme will preserve the order.
The iowait drain/wait mechanisms are extended for a pio case. An
additional qp wait flag is added for the PIO drain wait case.
Currently the only pio wait is for buffers, so the no_bufs_available()
routine name is changed to pio_wait() and a third argument is passed
with one of the two pio wait flags to generalize the routine. A module
parameter is added to hold a configurable threshold. For now, the
module parameter is zero.
A heuristic routine is added to return the func pointer of the proper
egress routine to use.
The heuristic is as follows:
- SMI always uses pio
- GSI,UD qps <= threshold use pio
- UD qps > threadhold use sdma
o No coordination with sdma is required because order is not required
and this qp pio count is not maintained for UD
- RC/UC ONLY packets <= threshold chose as follows:
o If sdmas pending, use SDMA
o Otherwise use pio and enable the pio tracking count at
the time the pio buffer is allocated
- RC/UC ONLY packets > threshold use SDMA
o If pio's are pending the pio_wait with the new wait flag is
called to delay for pios to drain
The threshold is potentially reduced by the QP's mtu.
The sc_buffer_alloc() has two additional args (a callback, a void *)
which are exploited by the RC/UC cases to pass a new complete routine
and a qp *.
When the shadow ring completes the credit associated with a packet,
the new complete routine is called. The verbs_pio_complete() will then
decrement the busy count and trigger any drain waiters in qp destroy
or reset.
Reviewed-by: Jubin John <jubin.john@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mike Marciniszyn [Sun, 14 Feb 2016 20:45:18 +0000 (12:45 -0800)]
staging/rdma/hfi1: fix panic in send engine
The send engine wasn't correctly handling
pre-built packets, and worse, the pointer to
a packet state's txreq wasn't initialized correctly.
To fix:
- all waiters need to save any prebuilt packets
(smda waits already did)
- the progress routine needs to handle a QPs prebuilt packet
and initialize the txreq pointer properly
To keep SDMA working, the dma send code needs to see if
a packet has been built already. If not the code will build
it.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Remove header memcpy from sdma send path.
Instead of writing the header into a buffer then copying it into another
buffer to be sent, remove that memcpy and instead build the header directly
into the tx request that will be sent.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Vennila Megavannan <vennila.megavannan@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mike Marciniszyn [Sun, 14 Feb 2016 20:44:34 +0000 (12:44 -0800)]
staging/rdma/hfi1: move txreq header code
The patch separates the txreq defines into new files, one for
verbs and one for sdma.
The verbs_txreq implementation handles the setup and teardown
of the txreq cache, so the register routine is changed to call
the new init/exit routines.
This patch allows for followup patches enhance the send engine.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Harish Chegondi [Sun, 14 Feb 2016 20:11:28 +0000 (12:11 -0800)]
IB/qib: Destroy SMI AH before de-allocating the protection domain
If SMI AH is not destroyed before de-allocating the PD, it would result in
non-zero PD use count when de-allocating the PD, triggering a WARN_ON() at
drivers/infiniband/core/verbs.c:284 ib_dealloc_pd+0x69/0xb0 [ib_core]()
when unloading the qib driver on systems with dual-port card.
This problem has always been there in qib and was detected only after the
commit 7dd78647a2c2 ("IB/core: Make ib_dealloc_pd return void") introduced
a WARN_ON in ib_dealloc_pd() that triggers if a PD's use count is non-zero
before de-allocating the PD.
Remove exported functions which are no longer required as the
functionality has moved into rdmavt. This also requires re-ordering some
of the functions since their prototype no longer appears in a header
file. Rather than add forward declarations it is just cleaner to
re-order some of the functions.
Reviewed-by: Jubin John <jubin.john@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Initially it was intended that rdmavt would support some signaling
between the underlying driver and itself. However this turned out to be
unnecessary for qib and hfi1. If we need to add something like this in
later to support another driver we should do it then. As of now this
essentially dead code so remove it.
Reviewed-by: Ira Weiny <ira.weiny@intel.com> Reviewed-by: Jubin John <jubin.john@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
While hfi1 and qib were still supporting bits and pieces of core verbs
components there needed to be a way to convey if rdmavt should handle
allocation and initialize of resources like the queue pair table. Now
that all of this is moved into rdmavt there is no need for these flags.
They are no longer used in the drivers.
Reviewed-by: Ira Weiny <ira.weiny@intel.com> Reviewed-by: Jubin John <jubin.john@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
IB/qib: Setup notify free/create mad agent callbacks for rdmavt
Qib needs to be notified when mad agents are created and freed, there is
some counter maintenance that needs to be performed. Add those callbacks at
registration time with rdmavt.
Reviewed-by: Ira Weiny <ira.weiny@intel.com> Reviewed-by: Jubin John <jubin.john@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
For each verb validate that all requirements for driver callbacks are met.
If a function is called without checking for a valid pointer, it is a
required function. Also document what each callback function does.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
IB/rdmavt: Clean up comments and add more documentation
Add, remove, and otherwise clean up existing comments that are leftover
from the initial code postings of rdmavt. Many of the comments were added
to provide an idea on the direction we were thinking of going. Now that the
design is solidified make a pass over and clean everything up. Also add
details where lacking.
Ensure all non static functions have nano comments.
Reviewed-by: Jubin John <jubin.john@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Kaike Wan [Sun, 14 Feb 2016 20:10:20 +0000 (12:10 -0800)]
staging/rdma/hfi1: Put QPs into error state after SL->SC table changes
If an SL->SC mapping table change occurs after an RC/UC QP is created,
there is no mechanism to change the SC nor the VL for that QP. The fix
is to place the QP into error state so that ULP can recreate the QP with
the new SL->SC mapping.
Reviewed-by: Ira Weiny <ira.weiny@intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Kaike Wan <kaike.wan@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mike Marciniszyn [Sun, 14 Feb 2016 20:10:04 +0000 (12:10 -0800)]
IB/qib, staging/rdma/hfi1: add s_hlock for use in post send
This patch adds an additional lock to reduce contention on the s_lock.
This lock is used in post_send() so that the post_send is not
serialized with the send engine and other send related processing.
To do this the s_next_psn is now maintained on post_send() while
post_send() related fields are moved to a new cache line. There is
an s_avail maintained for the post_send() to mitigate trading cache
lines with the send engine. The lock is released/acquired around
releasing the just built packet to the egress mechanism.
Reviewed-by: Jubin John <jubin.john@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Harish Chegondi <harish.chegondi@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Ira Weiny <ira.weiny@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Harish Chegondi [Sun, 14 Feb 2016 20:09:55 +0000 (12:09 -0800)]
IB/qib: Rename several functions by adding a "qib_" prefix
This would avoid conflict with the functions in hfi1 that have similar
names when both qib and hfi1 drivers are configured to be built into
the kernel. This issue came up in the 0-day build report.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Harish Chegondi <harish.chegondi@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
IB/rdmavt, staging/rdma/hfi1: use qps to dynamically scale timeout value
A busy_jiffies variable is maintained and updated when rc qps are
created and deleted. busy_jiffies is a scaled value of the number
of rc qps in the device. busy_jiffies is incremented every rc qp
scaling interval. busy_jiffies is added to the rc timeout
in add_retry_timer and mod_retry_timer. The rc qp scaling interval
is selected based on extensive performance evaluation of targeted
workloads.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Vennila Megavannan <vennila.megavannan@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: actually use new RNR timer API in loopback path
The patch series which added a new API for the RNR timer did not include an
updated call in the loopback path. RC/UC RNR loopback would be broken
without this.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Tune for unknown channel if configuration file is absent
Currently, the driver fails to tune the SerDes and therefore prevents
link up if the configuration file is missing or fails parsing or
validation. This patch adds a fallback option so that the 8051 is asked
to tune for an unknown channel and possibly get the link up if tuning
succeeds. It also adds a user-friendly message to update the
configuration file if it is out-of-date.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Fetch platform configuration data from EFI variable
The platform configuration data has been moved into the EFI variable
store where it is populated by the HFI1 option ROM. This patch pulls
the configuration data from the new location, retaining a fallback to
request_firmware.
Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
IB/qib: Insure last cursor is updated prior to complete
This patch is a prerequisite for adding a separate lock
for post send.
The timing of updating s_last needs to be before returning
any send completion to avoid a race between a poll cq seeing
a completion and the post send checking for a full queue.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Insure last cursor is updated prior to complete
This patch is a prerequisite for adding a separate lock
for post send.
The timing of updating s_last needs to be before returning
any send completion to avoid a race between a poll cq seeing
a completion and the post send checking for a full queue.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Adding support for hfi counters via sysfs
It enables access to counters in
/sys/class/infiniband/hfi1_0/ports/1/counters
by providing infrastructure when PMA queries occur. Counters symbol_error
and VL15_dropped are not supported in OPA, therefore, 0 will always be
returned. In addition, two common routines (pma_get_opa_port_dctrs,
pma_get_opa_port_ectrs) were created to query counters to avoid code
duplication.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mitko Haralanov [Wed, 3 Feb 2016 22:37:41 +0000 (14:37 -0800)]
staging/rdma/hfi1: Fix bug that could block the process on context exit
A race was discovred in the user SDMA code, which could result
in an process being stuck in the kernel call indefinitely in
certain error conditions.
If, during the processing of a user SDMA request, there was an
error *and* all outstanding SDMA descriptor had been completed
by the time the that error case was handled in the calling function,
the state of the packet queue would not get correctly updated
resulting in the process subsequently getting stuck, thinking that
there are more descriptors to be completed.
To handle this scenario, the driver now checks the submitted
packet count vs. the completed. If all submitted packets have also
been completed, the driver can safely free the request and signal
user level. Otherwise, this will be handled by the completion
callback.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Dean Luick [Wed, 3 Feb 2016 22:37:24 +0000 (14:37 -0800)]
staging/rdma/hfi1: Make EPROM check per device
Add a variable eprom_available to each device, replacing the
global of the same name. This is to allow multiple HFI devices
with different EPROM availability to operate correctly on the
the same system.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Add credits for VL0 to VL7 in snoop mode
Add a new option to the snoop ioctl which allows credits to be allocated
across all VLs. Previously only VL0 and VL15 had credits allocated.
The new option used in the ioctl HFI1_SNOOP_IOCSET_OPTS allows credits
to be allocated so that VL15 will have at least 8.5KB credits and the
other VLs will have the rest of the credits divided equally across
themselves.
The total number of credits are stored in the upper 16 bits of the
integer passed and the cumulative value should ensure that VL0 has at
least 8.5KB and each VL a minimum of 2KB + 128 bytes
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Sadanand Warrier <sadanand.warrier@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mitko Haralanov [Wed, 3 Feb 2016 22:37:06 +0000 (14:37 -0800)]
staging/rdma/hfi1: Improve performance of user SDMA
To facilitate locked page counting, the user SDMA
routines would maintain a list of io vectors, which
were freed in the completion callback and then unpin
the associated pages during the next call into the
kernel.
Since the size of this list was unbounded, doing this
was bad for performance because the driver ended up
spending too much time freeing the io vectors.
This commit changes how the io vector freeing is done
by moving the actual page unpinning in the callback and
maintaining a count of unpinned pages. This count can
then be used during the next call into the kernel to
update the mm->pinned_vm variable (since that requires
process context and the ability to sleep.)
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Harish Chegondi [Wed, 3 Feb 2016 22:36:49 +0000 (14:36 -0800)]
staging/rdma/hfi1: Remove modify_port and port_immutable functions
Delete code from query_port which has been moved into rvt_query_port
Create a call back function to shut down a port which may be called from
rvt_modify_port
Query gid is in rdmavt, but still relies on the driver to maintain the
guid table. Add the necessary driver call back and remove the existing
verb handler.
Jubin John [Wed, 3 Feb 2016 22:36:31 +0000 (14:36 -0800)]
staging/rdma/hfi1: Clean up init_cntrs()
Clean up init_cntrs() by removing unnecessary memsets and debug
statements
Suggested-by: Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
The LRH has a 12 bit packet length field, not 11 bit. This caused a
snoop packet length miscalculation leading to a crash when sending a
large ping over IPoIB while running opapacketcapture.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Implement LED beaconing for maintenance
This patch implements LED beaconing for maintenance. A MAD packet with
the LEDInfo attribute set to 1 will enable LED beaconing with a duty
cycle of 2s on and 1.5s off. A MAD packet with the LEDInfo attribute
set to 0 will disable beaconing and return the LED to normal operation.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Dean Luick [Wed, 3 Feb 2016 22:35:49 +0000 (14:35 -0800)]
staging/rdma/hfi1: Split last 8 bytes of copy to user buffer
Copy the last 8 bytes of user mode RC WRITE_ONLY and WRITE_LAST
opcodes separately from the rest of the data.
It is a de-facto standard for some MPI implementations to use a
poll on the last few bytes of a verbs message to indicate that
the message has been received rather than follow the required
function method. The driver uses the kernel memcpy routine, which
becomes "rep movsb" on modern machines. This copy, while very
fast, does not guarantee in-order copy completion and the result
is an occasional perceived corrupted packet. Avoid the issue by
splitting the last 8 bytes to copy from the verbs opcodes where it
matters and performing an in-order byte copy.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Dean Luick [Wed, 3 Feb 2016 22:35:40 +0000 (14:35 -0800)]
staging/rdma/hfi1: Fix fabric serdes reset by re-downloading firmware
A host fabric serdes reset is required to go back to polling.
However, access to the fabric serdes may have been invalidated
by the sibling HFI when it downloads its fabric serdes firmware.
Work around this by re-downloading and re-validating the serdes
firmware at reset time on Bx hardware.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mitko Haralanov [Wed, 3 Feb 2016 22:35:23 +0000 (14:35 -0800)]
staging/rdma/hfi1: Properly determine error status of SDMA slots
To ensure correct operation between the driver and PSM
with respect to managing the SDMA request ring, it is
important that the status for a particular request slot
is set at the correct time. Otherwise, PSM can get out
of sync with the driver, which could lead to hangs or
errors on new requests.
Properly determining of when to set the error status of
a SDMA slot depends on knowing exactly when the last txreq
for that request has been completed. This in turn requires
that the driver knows exactly how many requests have been
generated and how many of those requests have been successfully
submitted to the SDMA queue.
The previous implementation of the mid-layer SDMA API did not
provide a way for the caller of sdma_send_txlist() to know how
many of the txreqs in the input list have actually been submitted
without traversing the list and counting. Since sdma_send_txlist()
already traverses the list in order to process it, requiring
such traversal in the caller is completely unnecessary. Therefore,
it is much easier to enhance sdma_send_txlist() to return the
number of successfully submitted txreqs.
This, in turn, allows the caller to accurately determine the
progress of the SDMA request and, therefore, correctly set the
error status at the right time.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Dean Luick [Wed, 3 Feb 2016 22:35:14 +0000 (14:35 -0800)]
staging/rdma/hfi1: correctly check for post-interrupt packets
At the end of the packet processing interrupt and thread handler,
the RcvAvail interrupt is finally cleared down. There is a window
between the last packet check (via DMA to memory) and interrupt
clear-down. The code to recheck for a packet once the RcvAVail
interrupt is enabled must ultimately use a CSR read of RcvHdrTail
rather than depend on DMA'ed memory.
This change adds a CSR read of RcvHdrTail if the memory check does
not show a packet preset. The memory check is retained as a quick
test before doing the more expensive, but always correct, CSR read.
In the ASIC, the CSR read used to force the RcvAvail clear-down write
to complete may bypass queued DMA writes to memory. The only correct
way to decide if a packet has arrived without an interrupt to push DMA
to memory ahead of itself is to read the tail directly after RcvAvail
has been cleared down. It is not sufficient to just read the tail and
skip pushing the clear-down. Both must be done. The tail read will not
push clear-down write due to it being in a different area of the chip.
At this point, it is OK to have packet data still being DMA'ed to
memory. This is the end of packet processing for previous packets.
If the driver detects a new packet has arrived before interrputs were
re-enabled, it will force a new interrupt and the interrupt will push
the packet DMAs to memory, where the driver will then react to the
interrupt and do normal packet processing.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mitko Haralanov [Wed, 3 Feb 2016 22:35:06 +0000 (14:35 -0800)]
staging/rdma/hfi1: Improve performance of SDMA transfers
Commit a0d406934a46 ("staging/rdma/hfi1: Add page lock limit
check for SDMA requests") added a mechanism to
delay the clean-up of user SDMA requests in order to facilitate
proper locked page counting.
This delayed processing was done using a kernel workqueue, which
meant that a kernel thread would have to spin up and take CPU
cycles to do the clean-up.
This proved detrimental to performance because now there are two
execution threads (the kernel workqueue and the user process)
needing cycles on the same CPU.
Performance-wise, it is much better to do as much of the clean-up
as can be done in interrupt context (during the callback) and do
the remaining work in-line during subsequent calls of the user
process into the driver.
The changes required to implement the above also significantly
simplify the entire SDMA completion processing code and eliminate
a memory corruption causing the following observed crash:
This commit also addresses issues related to notification of user
processes of SDMA request slot availability. The slot should be
cleaned up first before the user processes is notified of its
availability.
Reviewed-by: Arthur Kepner <arthur.kepner@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Reduce syslog message severity and provide speed information
The syslog message causes unnecessary alarm for the single and dual port
x8 cards by reporting at an error level. This patch reduces the severity
to informational only and adds speed information.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mitko Haralanov [Wed, 3 Feb 2016 22:34:41 +0000 (14:34 -0800)]
staging/rdma/hfi1: Improve performance of TID cache look up
When TID caching was enabled, the way the driver found
RB nodes when PSM was unprogramming TID entries was by
traversing the RB tree, looking for a match on the
RcvArray entry index.
The performance of this algorithm was not only poor but
also inconsistent depending on how many RB nodes would
have to be traversed before a match was found.
The lower performance was especially evident in cases where
there was a cache miss with the cache full, requiring the
unprogramming of several TID entries.
This commit changes how RB nodes are looked up when being
free'd by PSM to a index-based lookup into a flat array on
the index of the RcvArray entry. This turns the entire
look-up process into an O(1) algorithm.
Special care needs to be taken for situations when TID
caching is disabled. In those cases, there is no need to
insert the RB nodes into an actual RB tree. Since the entire
RcvArray management mechanism is managed by an index-based
algorithm, the RB nodes can be saved into the flat array,
making both "insertion" and "removal" faster.
Reviewed-by: Arthur Kepner <arthur.kepner@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Fix for module parameter rcvhdrcnt when it's 2097152
The driver crashes when loaded with parameter rcvhdrcnt=2097152.
The root cause was that rcvhdrcnt was initially a 32 bit variable
and its value was assigned to a 16 bit variable, truncating the
upper 16 bits. This patch prevents the user from passing a value
for rcvhdrcnt greater than 16352 (Maximum number for rcvhdrcnt).
Reviewed-by: Dean Luick <dean.luick@intel.com> Reviewed-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
This patch fixes the fairness issues in QP scheduling
- the timeout for cond_resched is changed to a ratio of
qp->timeout_jiffies
- workqueue_congested is used to determine if qp needs to
reschedule itself
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Signed-off-by: Vennila Megavannan <vennila.megavannan@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Dean Luick [Wed, 3 Feb 2016 22:34:15 +0000 (14:34 -0800)]
staging/rdma/hfi1: Fix for generic I2C interface
The original I2C interface was geared for QSFP accesses. Modify
the interface to behave more like a generic I2C controller such
that reads and writes can accept multi-byte offsets. Removed
reads following writes and moved reset to top level.
Reviewed-by: Easwar Hariharan <easwar.hariharan@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Pablo Cacho <pablo.cacho@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mitko Haralanov [Wed, 3 Feb 2016 22:33:40 +0000 (14:33 -0800)]
staging/rdma/hfi1: Consolidate CPU/IRQ affinity support
This patch unifies the affinity support for CPU and IRQ allocations into
a single code base. The goal is to allow the driver to make intelligent
placement decision based on an overall view of processes and IRQs across
as much of the driver as possible.
Pulling all the scattered affinity code into a single code base lays the
ground work for accomplishing the above goal. For example, previous
implementations made user process placement decision solely based on
other user processes. This algorithm is limited as it did not take into
account IRQ placement and could result in overloading certain CPUs.
A single code base also provides a much easier way to maintain and debug
any performance issues related to affinity.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Signed-off-by: Mitko Haralanov <mitko.haralanov@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Ashutosh Dixit [Wed, 3 Feb 2016 22:33:06 +0000 (14:33 -0800)]
staging/rdma/hfi1: Add support for enabling/disabling PCIe ASPM
hfi1 HW has a high PCIe ASPM L1 exit latency and also advertises an
acceptable latency less than actual ASPM latencies. Additional
mechanisms than those provided by BIOS/OS are therefore required to
enable/disable ASPM for hfi1 to provide acceptable power/performance
trade offs. This patch adds this support.
By means of a module parameter ASPM can be either (a) always enabled
(power save mode) (b) always disabled (performance mode) (c)
enabled/disabled dynamically. The dynamic mode implements two
heuristics to alleviate possible problems with high ASPM L1 exit
latency. ASPM is normally enabled but is disabled if (a) there are any
active user space PSM contexts, or (b) for verbs, ASPM is disabled as
interrupt activity for a context starts to increase.
A few more points about the verbs implementation. In order to reduce
lock/cache contention between multiple verbs contexts, some processing
is done at the context layer before contending for device layer
locks. ASPM is disabled when two interrupts for a context happen
within 1 millisec. A timer is scheduled which will re-enable ASPM
after 1 second should the interrupt activity cease. Normally, every
interrupt, or interrupt-pair should push the timer out
further. However, since this might increase the processing load per
interrupt, pushing the timer out is postponed for half a second. If
after half a second we get two interrupts within 1 millisec the timer
is pushed out by another second.
Finally, the kernel ASPM API is not used in this patch. This is
because this patch does several non-standard things as SW workarounds
for HW issues. As mentioned above, it enables ASPM even when advertised
actual latencies are greater than acceptable latencies. Also, whereas
the kernel API only allows drivers to disable ASPM from driver probe,
this patch enables/disables ASPM directly from interrupt context. Due
to these reasons the kernel ASPM API was not used.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Signed-off-by: Ashutosh Dixit <ashutosh.dixit@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Mitko Haralanov [Wed, 3 Feb 2016 22:32:49 +0000 (14:32 -0800)]
staging/rdma/hfi1: Correctly set RcvCtxtCtrl register
The RcvCtxtCtrl register was being incorrectly set upon context
initialization and clean up resulting, in many cases, of contexts using
settings from previous contexts' initialization. This resulted in bad
and unexpected behavior. This was especially important for the TailUpd
bit, which requires special handling and if set incorrectly could lead
to severely degraded performance.
This patch fixes the handling of the RcvCtxtCtrl register, ensuring that
each context gets initialized with settings applicable only for that
context. It also ensures the proper setting for the TailUpd bit by
setting it to either 0 or 1 (as needed by the context's configuration)
explicitly.
staging/rdma/hfi1: Fix for 32-bit counter overflow in driver and hfi1stats
When 32-bit hardware counters overflow, hfi1stats misinterprets
the counters as being 64 bits causing the deltas for the
counters to be a huge number. This patch makes hfi1stats
aware that a counter is 32 bits by making the driver write
<counter name>,32 to debugfs.
Reviewed-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Dean Luick [Wed, 3 Feb 2016 22:32:23 +0000 (14:32 -0800)]
staging/rdma/hfi1: No firmware retry for simulation
Simulation has no firmware, so it will never move firmware
acquire to the FINAL state. Avoid that by skiping the TRY
state and moving directly to FINAL.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Don't attempt to qualify or tune loopback plugs
Loopback plugs used for testing hardware don't need to be qualified to
bring the link up unlike production cables. This patch adds an exception
for loopback plugs to the QSFP and SerDes tuning algortihm.
Implement per-VL transmit counters. Not all errors can be
attributed to a particular VL, so make a best attempt.
o Extend the egress error bits used to count toward transmit
discard.
o When an egress error or send error occur, try to map back
to a VL.
o Implement a SDMA engine to VL (back) map.
o Add per-VL port transmit counters
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
The gen3 bump code must mark a firmware download failure as fatal.
Otherwise a later load attempt will fail with a NULL dereference.
Also:
o Only do a firmware back-off for RTL. There are no alternates for
FPGA or simulation.
o Rearrange OS firmware request order to match what is actually
loaded. This results in more coherent informational messages
in the case of missing firmware.
Reviewed-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Dean Luick <dean.luick@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Support external device configuration requests from 8051
This patch implements support for turning on and off the clock data
recovery mechanisms implemented in QSFP cable on request by the DC 8051
on a per-lane basis.
staging/rdma/hfi1: Get port type from configuration file
The current code employs a heuristic to guess the port type.
The canonical location to identify the port type of the
designed platform is from the platform configuration data.
This patch uses the previously fetched port type from the platform
configuration and removes the now obsolete heuristic routine
and its associated defines.
staging/rdma/hfi1: Add active and optical cable support
This patch qualifies and tunes active and optical cables for optimal
bit error rate and signal integrity settings. These settings are
fetched from the platform configuration data.
Based on attributes of the QSFP cable as read from the SFF-8636
compliant memory map, we select the appropriate settings from the
platform configuration data (examples: TX/RX equalization, enabling
cable high power, enabling TX/RX clock data recovery mechanisms, and RX
amplitude control) and apply them to the SERDES and QSFP cable.
The platform configuration data also contains system parameters such
as maximum power dissipation supported, and the cables are qualified
based on these parameters. As part of qualifying the cables, the
correct OfflineDisabledReasons are set for the appropriate scenarios.
Reviewed-by: Dean Luick <dean.luick@intel.com> Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Brent R Rothermel <brent.r.rothermel@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
staging/rdma/hfi1: Fix QSFP memory read/write across 128 byte boundary
The QSFP memory cache reads both lower and upper page 0H in one shot,
which leads to the address counter wrapping around to the beginning of
lower page 00H at byte 128, as defined by SFF-8636.
This patch fixes this by modifying the underlying QSFP read and writes
to avoid this wrap around.
Reviewed-by: Dean Luick <dean.luick@intel.com> Reviewed-by: Ira Weiny <ira.weiny@intel.com> Signed-off-by: Easwar Hariharan <easwar.hariharan@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
Bryan Morgan [Wed, 3 Feb 2016 22:30:49 +0000 (14:30 -0800)]
staging/rdma/hfi1: HFI reports wrong offline disabled reason when cable removed
Removing QSFP cable should report 'No Local Media' instead of
'Transient' as reported by 'opaportinfo'.
Workaround is to change the state to
OPA_LINKDOWN_REASON_LOCAL_MEDIA_NOT_INSTALLED in cable handler.
With cable still removed, 'opaportinfo bounce' should not cause a
state change to Polling, as reported by 'opaportinfo'.
Resolution is to prevent physical state change from Offline->Polling.
Use a macro to mask lower nibble of OPA_LINKDOWN_REASON* as needed
for offline_disabled_reason.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com> Reviewed-by: Easwar Hariharan <easwar.hariharan@intel.com> Reviewed-by: Dean Luick <dean.luick@intel.com> Reported-by: Todd Rimmer <todd.rimmer@intel.com> Signed-off-by: Bryan Morgan <bryan.c.morgan@intel.com> Signed-off-by: Jubin John <jubin.john@intel.com> Signed-off-by: Doug Ledford <dledford@redhat.com>
projects / linux-beck.git / log