siginfo_t is a GNU extension. SPDK (and DPDK) have
direct dependencies on GNU extensions, but it's a bit
nicer if external modules don't also need to define
_GNU_SOURCE. Currently siginfo_t parameter in the
spdk_pci_error_handler is the only thing that violates
this.
Note that DPDK also supports registering sigbus handlers,
but they take the failing address as a parameter instead
of the full siginfo_t structure. Let's adopt the same
for SPDK.
While here, remove an extra semicolon that was just after
the virtio sigbus handler function signature that was
updated in this patch.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I07faf11a3ac3589c637cb2196581c102286b1e68
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8333
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jacek Kalwas <jacek.kalwas@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
The PCIE and VFIOUSER both can use this function, the only difference
is VFIOUSER should use IOVA=VA to do the vtophys translation, so
here we will move the function to the common PCIe layer as the first
step.
Change-Id: I699edb67a00a2fa534072fc02ac2dd4a27aba8f4
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8030
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Read CAP (Capabilities) register as part of controller
initialization instead of controller construction.
For now, still read CAP in the pcie and vfio-user
controller construction, since they need the
drstd (doorbell stride) to construct the admin
queue.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I000fe880f2ec0d6de1d565c883d7ea0ae1ac2c81
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8078
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Read VS (Version) register as part of controller
initialization instead of controller construction.
This prepares for upcoming changes to make
controller attach fully asynchronous. Since reading
fabrics registers is an asynchronous operation, it
will be easier to read the VS register as part of
controller initialization which operates as an
asynchronous state machine.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I771386dbdf5902633e0d9f91b3b20be98f26fdc3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8076
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
spdk_vtophys() takes a mapping_length parameter, so
it can return the length for which the returned
virtual address is valid.
But spdk_vtophys() will only return the max
between the valid length and the input mapping_length
parameter.
So the nvme SGL building code for contiguous buffers
was broken, since it would only set the mapping_length
once, before the loop started. Worst case, if a buffer
started just before (maybe 256 bytes) before a huge page
boundary, each time through the loop we would create
a new SGL for only 256 bytes at a time, very quickly
running out of SGL entries for a large buffer.
Fixes#1852.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ib1000d8b130e8e4bfeacccd6e60f8109428dfc1e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7659
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
NVME_MAX_PRP_LIST_ENTRIES has changed over time, so let's
just remove the reference to the exact value here. Also
explain a bit more why the max size isn't
(NUM_ENTRIES + 1) * page_size.
While here, do a small whitespace cleanup as well.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ib75813788abdd3dbb43192f9fdc27f99b33aeadf
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7328
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Check if qpair has a poll group during the connect process,
use poll group's statistics or allocate own structure per
qpair. That is done due to not all applications use poll
groups and we want to avoid "if (qpair->group)"
conditions in data path.
Admin qpair always allocates its own statistics
structure but the statistics are not reported
since this qpair is not attached to a poll group.
Statistics are reported by spdk_nvme_perf tool
if --transport-stats and in bdev_nvme_transport_statistics
RPC method.
Change-Id: I58765be161491fe394968ea65ea22db1478b219a
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6304
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
The hotplug lib can be used for pcie devices
such as nvme, virtio_blk and virtio scsi.
For the sigbus handler, there is only one in a
process and it should handle all the devices.
And align nvme to the hotplug lib
Add the ADD uevent support for allowing the
device hotplug.
Change-Id: I82cd3b4af38ca24cee8b041a215a85c4a69e60f7
Signed-off-by: Jin Yu <jin.yu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5653
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This function allows applications to specify whether
they wish to allow probing a newly attached NVMe
PCIe SSD.
The env layer will only even probe devices that have
been allowed. By default, this is all devices, but
if the user has specified some list of
allowed PCI addresses (via spdk_env_opts pci_allowed)
then newly attached PCIe devices are implicitly not
allowed. This API allows applications to add
device addresses to the allowed list after the
application has started.
This API will be useful for use cases where multiple
SPDK processes are running on one server, and assignment
of PCIe SSDs to those processes are based on some function
of the SSD's PCIe address.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I919bc267f2ad9130ab5c875ff760a301028b047e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6184
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: <dongx.yi@intel.com>
The data path for PCIe and vfio-user transports are almost
same too, so move the code from nvme_pcie.c to nvme_pcie_common.c,
so that these APIs can be reused by vfio_user.
No logic change for this patch.
Change-Id: I82f480bba3bae0ce35e2a98f29839081095f7d50
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6040
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Intel P55XX SSDs can support SGL feature but can't use Dataset Management
command with SGL format, so add a quirk here for now, if the limitation was
fixed in future, we can remvoe this. Also SPDK doesn't privoide scatter buffer
API for DSM, so using PRP with DSM is totally fine.
Change-Id: Ibe92f4deb5b8bc2077115f5b7244bc17be4f3b23
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5858
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
There are some common data structures and APIs in pcie transport
which can be used both for pcie and vfio-user transport, so move
the common code into a new header and source file.
No actual logic change just the code movement except remove the
static function declarations.
Change-Id: Ie9021e703a5780fdd6840f0e3cfea76a0017a811
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5923
Community-CI: Broadcom CI
Reviewed-by: sunshihao <sunshihao@huawei.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When a device is removed, we should use the remove_cb
that was specified when the device was originally probed
and attached, if one was set.
Also add a new spdk_nvme_ctrlr_set_remove_cb API. This
can be used for cases where a different remove_ctx is
desired than was specified for the probe call. This
also enables setting a remove_cb when using connect APIs
which do not have a way currently to provide a remove_cb.
This also requires fixing the bdev nvme module, which
was depending on the previously errant behavior.
Fixes issue #1715.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Id848b39040099ff7a21fe57ea6b194a8c25ae015
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5510
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
spdk_vtophys() already checks that, so we don't need
to check it in the NVMe driver again.
Change-Id: I74288ae8cab80e1be34583475fa02a3ae13e090c
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5166
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This patch removes the string from register component.
Removed are all instances in libs or hardcoded in apps.
Starting with this patch literal passed to register,
serves as name for the flag.
All instances of SPDK_LOG_* were replaced with just *
in lowercase.
No actual name change for flags occur in this patch.
Affected are SPDK_LOG_REGISTER_COMPONENT() and
SPDK_*LOG() macros.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I002b232fde57ecf9c6777726b181fc0341f1bb17
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4495
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Mellanox Build Bot
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Broadcom CI
After a submission queue is deleted, the device is supposed
to post completions for every command to the completion queue.
Previously, we never looked and completed all commands with
an ABORTED status. Instead, complete any commands in the
completion queue with the status the drive gave them.
Change-Id: If851a365d4f305cf4390454b6b26dd0f7c5b82ac
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3875
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Registration macro now generates function based on driver's name.
It allows to have multiple registration within single source file.
Similar pattern is used e.g. by SPDK_NVMF_TRANSPORT_REGISTER.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: Ied0887e8dae7fe9ca1517313be5eff8f218b7e98
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3895
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
pctrlr->cmb.mem_register_addr and pctrlr->cmb.mem_register_size
are assigned after spdk_mem_register.
if spdk_mem_register is failed , ctrlr_map_cmb hasn't been executed.
they are not be used.
So remove them.
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Change-Id: I3d1996eee8b5260b79c4c3e0a2e1d376da2343b7
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3856
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Now that drivers can be registered from upper layers there's
no need to keep them centralized inside env.
(check_format.sh complains that spdk_pci_nvme_get_driver() shouldn't
start with the spdk_ prefix - to workaround that we move the function
declaration from one place in env.h to another - that's enough to
convince check_format it really is a public function)
Change-Id: If86aebd6c997349569c71430ec815b413eb44ef8
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3187
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jacek Kalwas <jacek.kalwas@intel.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is an attempt to workaround a kernel data race spotted in #1275.
We have to delay our VFIO interraction until after the kernel handles
the PCIe hotremoval. Otherwise we risk having a deadlock inside the
kernel. We could put a dummy delay in SPDK hotremove routine, but we
don't know exactly how long we have to wait, so wait for the VFIO
hotremove notification to arrive instead. From what I understand it
is always sent after the kerner already releases the critical locks.
This patch breaks hotremove for VFIO devices when running with
DPDK version < 18.05, but those reached EOL already and their support
is removed from SPDK later within this series.
It would be nice to wait for this VFIO notification inside lib/env_dpdk
e.g. inside the detach function, but we don't know there why the
detach is happening. It could be triggered by a user RPC, not physical
device hotremoval.
Change-Id: I2727de923bb2e3e05a83202465330a32a77d7cfc
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1732
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
To abort requests whose cb_arg matches, add child abort request greedily.
Iterating all outstanding requests is unique for each transport but
adding child abort is common among transports, and adding child abort
is replaceable by other operations.
Hence add qpair_iterate_requests() function to the function pointer table
of transport, and pass the operation done in the iteration by a
parameter of it.
In each transport, the implementation of qpair_iterate_requests() uses
TAILQ_FOREACH_SAFE() for potential future use cases.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic70d1bf2613fce2566eade26335ceed731f66a89
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2038
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Michael Haeuptle <michaelhaeuptle@gmail.com>
QEMU won't emulate wide (>8 byte) instructions on an MMIO
region. We tried to fix this previously but it wasn't sufficient.
So this version adds an explicit copy function that will only
copy 8 bytes at a time from the source cmd pointer into the
SQ cmd point in the virtualized CMB.
Fixes issue #1458.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I804005a2e8cf3f7365c1608674f82a3f843e59ba
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3135
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Each request has a callback context as cb_arg, and the callback to
nvme_complete_request() for the completed request may reuse the context
to the new request.
On the other hand, nvme_pcie_qpair_complete_tracker() dequeues tr from
pqpair->outstanding_tr after calling nvme_complete_request() for the
request pointed by tr.
Hence while nvme_complete_request() is executed, pqpair->outstanding_tr
may have two requests which has the same callback context, the
completed request and the new submitted request.
The upcoming patch will search all requests whose cb_arg matches to
abort them. In the above case, the search may find two requests by
mistake.
To avoid such error, change nvme_pcie_qpair_complete_tracker() to
dequeue tr from pqpair->outstanding_tr before calling nvme_complete_request().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ie9efc200d06d02d8ee2be3cb8e9fd64591bc210d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2861
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Michael Haeuptle <michaelhaeuptle@gmail.com>
This helps ensure thread safety on creation of the
netlink socket, when probe is called from multiple
threads at once. It is also a lot more clean - we just
create it once, rather than checking every time probe
is called to see if it has to be created.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I528cedc3ff44de6ea8ecaf6d2389226502ba408e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2681
Community-CI: Broadcom CI
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
remove only the spdk_ prefix from static functions in
the above libraries.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Change-Id: I59ce032c3312fa73f30c133fd62e603c1eee2859
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2365
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When starting the QEMU NVMe SSD with "cmb_size_mb=XX"
parameter, the controller memory buffer feature is
enabled in the Guest, the SPDK NVMe driver running
in the Guest will allocate the submission queue in
the controller memory buffer by default, it will use
memset and SSE instruction when copying NVMe command
to the submission queue entry or zero the whole
submission queue, inside the memset implementation
the AVX2 instuction will be used if the CPU can support
such feature. However, due to the limitation in the
QEMU, the maximum access width to the PCI BAR space is
set to 8 Bytes, SPDK will report illegal instruction
in Guest.
Here we add a quirk for the QEMU emulated NVMe.
Fix issue #1362.
Change-Id: Ib5e2e4198d39ce7f8455175f7db283db0b01eebf
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2196
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Following patch made sure that CQ/SQ are allocated in
physically contiguous manner:
(64db67) nvme/pcie: make sure sq and cq are physically contiguous
Using MAX_IO_QUEUE_ENTRIES is enough to make sure that either
queue does not span multiple hugepages.
Yet the patch made sure that whole page is occupied only
by the queue. Which unnecessarily increases memory consumption
up to two hugepages per each qpair.
This patch changes it so that each queue alignment is limited
up to its size.
Changes in hugepages consumed when allocating io_qpair in hello_world
application:
io_queue_size Without patch With patch
256 8MiB 0MiB
1024 12MiB 4MiB
4096 24MiB 16MiB
Note: 0MiB means no new hugepages were required and qpair fits into
previously allocated hugepages (see all steps before io_qpair
allocation in hello_world).
Intersting result of this patch is that since we required alignment
up to the hugepage size this resulted in reserving even two 2MiB
hugepages to account for DPDK internal malloc trailing element.
See alloc_sz in try_expand_heap_primary() within malloc_heap.c
This patch not only reduces overall memory reserved for the
queues, but decreases increase in heap consumption on DPDK side.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I75bf86e93674b4822d8204df3fb99458dec61e9c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2244
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This makes more sense within the context of the nvme driver and
helps us avoid the awkward situation of getting a failed_qp callback
on a qpair that simply hasn't been connected.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Change-Id: Ibac83c87c514ddcf7bd360af10fab462ae011112
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1734
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Commit 9ad044c4 remove the offset parameter in nvme_pcie_ctrlr_alloc_cmb()
and uses spdk_vtophys() to get the physical address of the submission
queue, however, the memory isn't registered to SPDK yet, so use the
offset to get the physical address again.
Change-Id: Ida0f4a8bf1e72d6446fba442a1b71f0ddaab4bb8
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1933
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
When iterating SGL segment, we can use existing virt_addr parameter
to return UINT64_MAX as a special value to indicate this segment need
to be described as Bit Bucket SGL.
Currently only READ command is supported, we can enable the WRITE
and COMPARE support when necessary.
Change-Id: I50aa2b226ec3449c13ed1d97b3224ee8e7de95a8
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1467
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
If the request contains separate metadata buffer, we should set
PSDT 10b when hardware SGL is supported and dword alignment is
required. SPDK driver doesn't provide SGL metadata data buffer
APIs for now, the separate metadata buffer is alway contiguous,
so we can put the metadata buffer into a data block SGL entry.
Change-Id: I9ccfce755a3169cd2c9b908e4da76081155d9613
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1466
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
alloc functions
Make the signature look more like spdk_dma_malloc and family.
Change-Id: I6a31acf3b9edd136924b6e53b9a23c699429b9dc
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/788
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Separate these two operations into different functions. It is
possible that a CMB may not be visible from the CPU, but still
be present and have data transferred to it by some other DMA
engine. Generalize the API to handle that case.
Change-Id: Ifcd282af0db734fe4a6ef2283ae8e8933d017809
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/787
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
We only need to register the memory for the CMB if the user
plans to map it for use with data. Delay the operation until then.
This also gives us a way to support unmapping the CMB.
Change-Id: I6121ae2cc3f6f44efae8b52b6582ff9f68432bc8
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/786
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Instead of creating an allocator where the driver manages the space,
now, since using the CMB for queues and data has already been
disallowed, just create functions to map and unmap the entire CMB.
The user can manage the space.
Change-Id: I023994deda3b517e14d2ba464c7375bf22b58456
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/785
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
Currently nvme_completion_poll_status object is allocated using
malloc, so it may cotnain some garbage. In some scenarious
nvme_completion_poll_cb can be triggered before we enter
spdk_nvme_wait_for_completion_*. In that case status object
will be freed by nvme_completion_poll_cb if it contains a
garbage in `timed_out` field. Later spdk_nvme_wait_for_completion
will work with already freed memory.
Fix - allocate nvme_completion_poll_status object using
calloc and explicitly zerofy it before usage
Fixes#1292
Change-Id: Iac39653a6cd102471de16e65814f0760bbeda7d9
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1373
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This API will allow us to simplify the polling mechanism for qpairs on a single
thread. It also will pave the way for doing transport specific aggregation of
qpair polling to increase performance.
The generic implementation is included. The transport specific calls
have yet to be implemented.
Change-Id: If07b4170b2be61e4690847c993ec3bde9560b0f0
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/579
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
When the drives report that SGL dword alignment is required in
Identify Controller data structure, when using separate metadata,
PSDT should only be set with 10b. The specification says: If PSDT
01b was used, Metadata Pointer (MPTR) contains an address of a
single contiguous physical buffer that is byte aligned.
For supporting this case, SPDK driver needs a metadata SGL entry,
so we can reserve one entry in the tracker data structure.
Change-Id: I2d86a58b0395c3000626f922e56d7f2212c8a752
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1316
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
PSDT 00b also need to check the metadta alignment.
Change-Id: I117f524c61bc4c712b46c91e4d51549825d06f6c
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1353
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
We only set the flag to false when the controller reports SGL supported
and can use byte contiguous buffer. Also check the data block's alignment
for hardware SGL.
Change-Id: Id936c49823963000d0543fc95fbb6edba3118feb
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1352
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
