Unit tests will be added as part of a separate patch updating all UT for
nvme.c. Global used for timeout value so it can be easily overwritten
by the upcoming unit tests for this function.
Change-Id: I7fc15aab91601ac57c94cae266b212c0998d2495
Signed-off-by: paul luse <paul.e.luse@intel.com>
The VirtualBox emulated NVMe device will intermittently
hang on the first read/write command after an I/O
qpair has been allocated. The frequency of the hang
diminishes if a delay is added after allocating the I/O
qpair - until it disappears completely with a 100us delay.
So add a quirk to insert this delay.
Note - the 100us delay was tested by running
the hello_world example app 50000 times.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I237e31b1b8a1a1e28262851ae0a21cd7345f0f1a
Fixes a scan-build warning about using qpairs after they have been
freed.
Change-Id: I263eabd6b784acf540c66136965f7705ef110a78
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Fix up the existing comment blocks misaligned in the first column.
Also add line numbers to the comment checks.
Change-Id: I9d28c365271df36e7013d74cbb02d0023ab4f581
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
A 100us is so small that applying the quirk to the specific
SSDs that require the delay is more trouble than it is worth.
So remove the quirk and always wait 100us before re-enabling
the NVMe SSD during initialization.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Id6a8cc6e35d103fffdf135580301fc3e5b27e722
Also avoid an spdk_get_ticks() call in the default
case where a timeout_cb_fn is not defined.
On my Intel(R) Xeon(R) E5-2699 v3 system with an
Intel(R) P3700 SSD, these modifications reduce software
overhead per I/O by 3-5% (as measured by the SPDK
overhead tool).
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I5bb5a87747b15d9e27655fabcd2bc1a40b0b990e
Queue aborts that would exceed the abort command limit
in software as a convenience for the user.
Change-Id: I8c1f0380984cc6c0cdb453db961939a7f571b336
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Instead, pass NULL when an ADMIN command times out.
We don't expose the admin queue to the user.
Change-Id: If0768d329a689f6f7c3734c9d419e680d7378ed1
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
For each command that times out, call the timeout
callback one time if the user registered one.
Change-Id: Iaad39a886468e89bef63fe292c5cad1dce97a57c
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
It has been discovered that some devices require
a very small delay before writing CC.EN to 1 after
CSTS.RDY goes to 0.
Change-Id: I73d31726d17ebf5bbec7ee528e2f98fcd05234dd
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This isn't the indentation pattern I would have chosen, but
it's a complicated negotiation between what I want and what
astyle will let me get away with.
Change-Id: I4909587823931842ac3f227134e1d05e7d80da74
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Any Intel device reporting device ID 0x0953 needs this quirk.
Change-Id: I690b01ecf05105df00ec8cf6f2da7f7c0a601aa8
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This prevents a deadlock if the user immediately
calls spdk_nvme_detach.
Change-Id: I79f28abe163cbbf184bea907692c44aa4e1c8893
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
The message about the uevent socket is not a fatal error; it just means
that hotplug monitoring will not work.
Change-Id: I29f6a253e96a86420c0fde9e19135f9f1d229bb9
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
remove the unnecessary rte_eal_pci_probe_one() in function
spdk_pci_device_detach(), this could cause error message when we
terminate the application, it will also not make sense try to probe one
device after we detach it, we could call spdk_pci_nvme_device_attach()
instead of spdk_pci_nvme_enumerate() when we have one given device address,
dpdk will try to scan the device and add it back to pci device list then.
Change-Id: I35f5bb412249bb20da57394f0531c10a49691906
Signed-off-by: Cunyin Chang <cunyin.chang@intel.com>
Register all spdk_malloc() memory regions as ibv_mr in a spdk_mem_map
so we can look up the RDMA key for the user's buffer and pass it in the SGL
directly, rather than copying through a pre-registered bounce buffer.
Change-Id: I7340bc2020b5256750c95dbd24ba67961404e5e7
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The extended LBA format flag should be initialized after namespace
capability flag.
Change-Id: Iad479b454bb4e31120c17d40ae23937a099c6f8f
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
The send completions must be processed prior to the
recv completions. However, if the completion queues
are separate this leaves a small window where
a send+recv completion arrive between polling
the send_cq and the recv_cq, resulting in the code
seeing the recv completion prior to the send
completion.
By combining the completion queues, this eliminates
any potential gap. The send completion will always
be processed before the recv completion.
Change-Id: I06bfef6af48559d0b9e00524ebc10f1a102e7387
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
The actual uses of intrinsics are already guarded by feature-specific
ifdefs in nvme_pcie_copy_command(), but the header itself should also
only be included when it will actually be needed.
Change-Id: Ife65d6432b8dfd9d9db80fe4e385ab76491874c0
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
These allocations need to be from memory registered with the SPDK env
library to allow future work on automatic ibverbs memory registration.
Change-Id: I6ec6999ecd6d6bf6ba4ab159630f7d01f3d46154
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This can be used for issuing an abort for the timed-out command.
Change-Id: I3c5727fdddc156cd7c8f99afbc3e6da8e73bba56
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Move the ibv_recv_wr initialization in
nvme_rdma_alloc_rsps. Thus we can save some
CPU times
Change-Id: Id449b2684290431f8b3ba97ec4058171d34038bf
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
We do not need to set it for submission since the contents
are same
Change-Id: I345094e2e8a858b318be73d28f09393566587d95
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Move cb_arg to the first argument to match the other NVMe callback
function signatures.
Change-Id: I4e699c8071dcb7ba4ce3cdb82ee985600208204c
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This has been reported for a number of different device
types. We suspect these devices are technically out of
spec, but they work with most other available NVMe
drivers on accident.
Change-Id: I529cfc03fc314cbab2a1cd40620bf1dd5b54182d
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
For namespaces with end-to-end protection information, metadata size
of exactly 8 bytes, and extended LBA configured, the NVMe driver would
calculate the size of the data block incorrectly. The NVMe spec has a
special provision for this specific case (8-byte metadata only) and
PRACT = 1 that requires that the host does not send the metadata as part
of the host memory buffer.
To fix this, clean up the calculation of the per-block data transfer
size by adding a new extended_lba_size field in the namespace, which
represents the total size of data to be transferred per block based on
the namespace's configured metadata size and whether it transfers
metadata as part of the data buffer. Then add the special case for
PRACT = 1 and PI configured and extended LBA in the R/W helper
functions.
Change-Id: I0b383a58c773cac06e6c018858b57129064c6059
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
These were repeated a few different places, so pull them into a common
header file.
Change-Id: Id807fa2cfec0de2e0363aeb081510fb801781985
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This removes one addition from the submission path (negligible, but a
nice side effect), but also opens up the possibility of reporting the
total time an I/O took - since we are always tracking the submission
time anyway, there is no extra cost to report it in the completion
callback.
Change-Id: I7129e7c09d20da8082042a7622d045846461dd9c
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
For infinite timeout states, instead of printing UINT64_MAX as a
decimal number, interpret it as "no timeout" instead.
Change-Id: I579f5857f96286734940ab5f493261e60354c4fe
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The PCIe transport initializes the quirks directly, so the generic hook
to get PCI ID is no longer necessary. This path was dead code.
Change-Id: I25bdaa598db53e4312a264d9d8356d1b416696e5
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The logic to fail queue pairs when the controller is failed should be
handled in the generic code, not in the individual transports.
This also allows nvme_qpair_fail() to be private to nvme_qpair.c.
Change-Id: I6194576dceb35073b9af8847e59314900028637c
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This is more CPU efficient than only grabbing one
completion per call to ibv_poll_cq.
Change-Id: I0c70d33639f0f345482d9e7c810f9c6723937058
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
In our previous code, we did not ack the event in
exceptional cases when we get a event via rdma_get_cm_event.
Thus, the code may block with in this statement:
rdma_destroy_id(rqpair->cm_id);
in some exceptiaonal cases. And this patch will solve this
issue.
Change-Id: Iddb6fb5356a5ee0ed04e261a040ba53042fca302
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
This make sure the qpair failure could be started from upper level application.
Change-Id: I7e04fe36929cc634ddf0078db96fbc40afb38f8c
Signed-off-by: Cunyin Chang <cunyin.chang@intel.com>
The g_thread_mmio_ctrlr should be not NULL pointer when it enter the
handler function.
Change-Id: I45dba601c672b16e2c6feafd9059bafde0d8f1b4
Signed-off-by: Cunyin Chang <cunyin.chang@intel.com>
If namespace is formatted with per lba metadata feature and also disable end-to-end protection
feature, host couldn't use per extended-lba metadata area.
Signed-off-by: Zhihao Zhang <thomas.zzh@alibaba-inc.com>
If the user asked for a specific PCI address in spdk_nvme_probe(), we
need to return 1, not 0, for the other PCI addresses that don't match
when enumerating. 0 means to attach the PCI driver, whereas 1 means to
continue enumerating.
With the previous behavior of returning 0, all NVMe devices would be
attached to the DPDK PCI driver, even if the user did not request for
them to be probed, and further calls to spdk_nvme_probe() would not find
any devices.
Change-Id: Ifbbcd7d1abe8ab535b6957855172e66a3e69fbe4
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Currently we use the pci functions provided by DPDK,
it identifies the device by class id related
info but not by pci bdf info, so we can add the filering
by pci_addr in pcie_nvme_enum_cb function.
Change-Id: I5942e98853f00fc10fa6aae5c113517653d1b357
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Since nvme_ns_cmd.c now walks the SGL, some of the test code
needs to also be updated to initialize and return correct values
such as ctrlr->flags and sge_length.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I521213695def35d0897aabf57a0638a6c347632e
Preivously, we only supports probe the NVMf target
via discovery info, now we can support to directly
to connect it.
Change-Id: I08ce1d95de6744286357e68b48c97b773b902ac8
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
(1) Add nvme_rdma_build_sgl_request function
(2) Merge nvme_rdma_pre/post_copy_mem to nvme_rdma_copy_mem
Change-Id: I86abab821b32b4da0aa9489a6b9f7dc430333159
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Now that the hotplug code is isolated in nvme_pcie.c, it can call the
PCIe transport attach function directly.
Change-Id: I2df3b9168473b537cc9b13367e06d3d3b6fa22be
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The spdk_nvme_qpair::num_entries value is never used in the common code,
so move it to the individual transport qpairs to make it clear that it
is a transport-specific implementation detail.
Change-Id: I5c8f0de4fcd808912ba6d248cf5cee816079fd32
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The definition of SPDK_UNREACHABLE uses the build-time DEBUG definition,
which is not available in the public API.
Change-Id: I1862c99fa5c85ccd3483f94e9c35de531da57f3c
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Instead of passing the work completion, just pass the
response index. This keeps the work completions localized
to the polling function.
Change-Id: I0e6a1d8564200b5ac3aa43dfd58ae152d439bbd8
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This eliminates an if statement, since the two callers
of this function know the desired queue size.
Change-Id: I28fabac8613f7b8fc7d96cf95b085b6e4dcf985f
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Just call the regular qpair create function instead.
Change-Id: Ic35b1eb6fcdf0d82733ea573a493f583dd63d5bd
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Use the num_entries value in the generic qpair instead. These
values had to match anyway.
Change-Id: Ia6400fbaba97df3ef6db4dc07a2ab95af1e5143f
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Avoid an extra level of pointer chasing when we are filling out the NVMe
SGL.
Change-Id: I1a40af16fda80f7480c419524876bfb1a1902eb8
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This will allow it to be better be reused for some future patches
enabling splitting of non-PRP-compliant SGL-based requests.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ica38fd6cf191f72baa524bcc4896b3c9939ab762
This intermediate function is no longer needed.
Change-Id: I3523cc6d8f3b290165a953d42cca8b76eda762c5
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Sending the fabric connect command is part of establishing
a connection, so move it into the main connection-establishing
function.
Change-Id: I55e7ffdd16b576c81b51d7d3910203f9afc1f4c2
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This function initializes the members of an existing
qpair struct. It doesn't construct one from scratch.
Change-Id: I0b9afac1ad25cfb217efd146702f693c74f5f697
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
No need to allocate all of the requests and responses until
we know a connection can be established.
Change-Id: I072a10aadfd7ced773634448f7d7e788622d0a4c
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
The code is clearer if this function is incorporated
into its only caller.
Change-Id: I33901cddf80ae27896b2acfd1b9e7d212f21f5f3
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This is resolving the address and route to the target, not
binding a socket to an address.
Change-Id: I80055481ed2e020410a1e186a4e7371b60faaee9
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
They were very close to the same already, so finish the job.
Change-Id: Ifba9e3b2d11a3e70cbfbe46f57a67552db2757ed
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
We should be sending the bounce buffer's remote key to the target so it
can put it into an RDMA SGE on the remote side.
Change-Id: Icded155ad2292c67baa722f001c9c07178bc2754
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
There is no particular reason for this to be 127; make it 128 to at
least be consistent with the PCIe transport.
Change-Id: I60500e0044d3549ba6350e1f35f09d624848bd21
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This function was only called from one place and saved no
lines of code.
Change-Id: If5e653732df57c1f2c93e20cf4f286eac31df91c
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This matches the behavior of nvme_pcie, which queues a request if no
tracker is available.
Change-Id: Idbf6c951c89451cfea22ec6bc553ff46f988f818
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Make the caller pre-allocate an rdma_req and change req_init() so it
only does initialization, not allocation.
This is necessary to distinguish between rdma_req allocation failure and
other types of failures, which will become important in future patches
when requests will be queued if rdma_req allocation fails.
Change-Id: Ie6edebc1b5f05001b42fc959a29ce0ea6875e41e
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Simplify the control flow and match the name of the function to its
purpose.
Change-Id: I65bad7e3b2ef710ca29eff9799b8dcaae3999315
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Make the qpair construct functions private to the transports - it
doesn't need to be called from generic code.
Change-Id: I5f730a4bcf60ce231fe27bc8f4c3c39cb647dd2d
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Add a transport callback to return the maximum queue size, and enforce
it in the generic nvme_ctrlr layer.
This allows the user to tell what io_queue_size was actually selected by
the transport via the ctrlr_opts returned during attach_cb.
Change-Id: I8a51332cc01c6655e2a3a171bb92877fe48ea267
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Equivalent to commit 6ab28a201b except now
for commands instead of responses.
Change-Id: Ibe4382dc0f65c1b90c2cee2ad285bbdd21b96a89
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The only field of bb_sgl that was actually used is lkey, and that is
already stored in bb_mr.
Change-Id: I790369a06ce223f88e356df20a9d9a74a93ff225
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Refactor the discover log page processing into a loop that calls a
function for each log page entry. This sets us up to add support for
multiple Get Log Page calls to handle larger discovery service lists.
Change-Id: I85676ada375d0dadda2a3f4ab6331123ac7aaf60
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This isn't used yet in the NVMe library, but it will be necessary later
for supporting non-IPv4 addresses.
Change-Id: I167ce63ad25b0e0c9aa192b12d764c8d078e67f9
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This better describes what the field controls (it does not affect the
admin queue size).
Change-Id: I851ae46fb4ed0fce819af07ae235824e0fc817e6
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
For requests split in _nvme_ns_cmd_split_request() the payload offset is
set after children are created using recurrent call _nvme_ns_cmd_rw().
This makes impossible to reset SGL to proper offset in
incomming patches that split non-PRP complaint SGL requests.
To change this the payload offset is set after each request is allocated
in _nvme_ns_cmd_rw() not in _nvme_ns_cmd_split_request().
Change-Id: I9d3b2e3bbd9d93a4c8a37e1db8c4e01276e2cacb
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
This is preparation for handling non-PRP compliant SGL.
Change-Id: I445790f9802292971256cf821d9730814c95a073
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
This is preparation for handling non-PRP compliant SGL.
Change-Id: I49c3745498411c5ff9e17cd08f181d4d434c2d08
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Don't leave garbage from previous discover entries in the trid we are
returning to the user.
Change-Id: I60ae5932db4a95cedb8df1ff98a2479220b55ce4
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The other simplifications to probe_info and trid made the
trtype argument redundant.
Change-Id: Ie7bea4e2204e690dc4909eeacd065e0722b53272
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
The probe_info was reduced to just containing a
transport_id, so remove probe_info entirely.
Change-Id: Ica9a22d126cd14e282decd3eea1a0afe0460f099
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This can be obtained by parsing traddr into a pci_addr,
then getting a handle to the pci_dev and asking for all
of the pci information.
Change-Id: I1948cbd3ec65611293192ef5558ace19dd444d4c
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This bug was preventing multiple calls to spdk_nvme_probe() from
working, since the first call would return 0 from all of the DPDK driver
init callbacks and prevent other devices from ever being enumerated in
subsequent calls.
Reported-by: Tsuyoshi Uchida <tuchida@us.fujitsu.com>
Change-Id: I871aa170bbd03be111604eeabe3a7a7a4f40ce89
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Use the standard quirk mechanism to specify which devices
need software assisted striping.
Change-Id: Id8156876a90b4caf9d687637e14c7ad4a66ceda6
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This way, all new controllers discovered will be initialized
in parallel.
Change-Id: Iebedb3905eb2787a3708f74425afae40ca31253d
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
If the first call to spdk_nvme_probe probes a device and
the driver elects not to take it, still call the probe
callback for that device on subsequence calls to
spdk_nvme_probe.
Change-Id: If06467cf6796c827a0bbfba6e36d5b91534526fc
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Move this down a level so it happens on all paths.
Change-Id: Iea9913f0e102353882466c8dea4ee39abb857520
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Scanning the transport may result in both new
devices and removed devices, so pass the callback
for both operations.
Change-Id: I6f73dbe6fd7cf61575c354b43f8ae3e2a01e2965
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Simplify the arguments to nvme_transport_ctrlr_scan to take
a transport id that identifies the discovery service (or
NULL to scan PCIe).
Further, separate scan into two functions - scan and attach.
Scan is for scanning an entire bus, attach is for a specific
device.
Change-Id: I464f351a02a04bc5a45096dcf5dc8fc5ac489041
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Instead of repeating the fields, just embed a transport_id.
Change-Id: I282704c9d59784abd5f7c93be4e47c673fcf6dde
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
This is a small step toward making discovery more like
scanning a local PCI bus.
Change-Id: Ie7149ad060f2eeb56939b1241187bdf09681f2aa
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Before adding readv/writev support in nvme_rdma,
using this patch.
Change-Id: I25ff0df61d0346f22560d011158d7f80e72007ea
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
NUMDU was added with NVMe 1.2.1 and allows a larger log page size to be
described.
Change-Id: I1a4ac42393c1a21175b3564980d56b6e7a6ae80d
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The NVMe over Fabrics transports should already be setting this in the
initial admin queue Connect command, so setting it again is not useful.
The kernel NVMe over Fabrics target additionally has a bug in the Set
Features - Keep Alive Timeout handler (it is extracting the KATO value
from the wrong offset in the command), so this works around the kernel
bug by not sending the Set Features command at all.
Change-Id: I0d7f09b71fcea116acf8810c5880157bb9315a04
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
The reason is that kernel nvmf target will check the
value. If not set, it will fail the other commands later.
Even for discovery ctrlr, kernel nvmf target will
check the cc value.
Change-Id: I998327f91ba96281d261952878eb84d648a823da
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
It's not the whole transport - it's just an enum for the
type of transport.
Change-Id: Ia435a21792f221ddf50ddf4f0923c6152622eccb
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change it according to the spec thus we can test
kernel nvmf target
Change-Id: Ica98dd40503a40c0f0de8efaefb1f6f67a89cde8
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change the PCI enumeration API to individual functions per device type
so that only the drivers that are actually in use get linked into the
final executable. All of the common code is still shared internally in
the env_dpdk library.
Change-Id: I2ba83afe59202a510f999a0674e23e60b6581221
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The user's remove_cb should detach the NVMe controller when it can
ensure that it is no longer in use. In the interim (between remove_cb
and spdk_nvme_detach()), the controller will remain in a failed state,
so any new I/O submissions will return an error code but not crash.
examples/nvme/hotplug is not yet updated for this change, but that will
be done in a separate patch.
Change-Id: I8827ba36f9688ccb734e7871f20f11ec11e88f96
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
While we're here, fix up typos and add error logs for all error exits
in nvme_rdma_qpair_connect().
Change-Id: I236fe6571c2012ca047aa8a447638d9227454c2f
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This version of multi-process support needs to have DPDK 16.11 builtin.
Change-Id: I3352944516f327800b4bd640347afc6127d82ed4
Signed-off-by: GangCao <gang.cao@intel.com>
The discover and probe 'nqn' fields are subsystem NQNs, so name them
subnqn to be consistent with the spec and the rest of the code and to
distinguish them from host NQNs.
Change-Id: I4a80fbc1f4b037c8a4f91c8f28d2a96e47c66c47
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Allow the host NQN to be overriden when connecting to NVMe over Fabrics
controllers.
Change-Id: I8fcf2e89ae7d9722677e834f76a8fe805c52f91b
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This makes the function and file/line info actually useful (instead of
pointing to the helper function itself).
Change-Id: I22bac68827115880a49d456706a7eaecdc12e9b5
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Each transport should handle its own qpair cleanup internally.
Change-Id: I7dd737be820ea6bad686f4aad7d74044fad58a47
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Let the transport access the controller options during
ctrlr_construct().
Change-Id: I83590c111e75c843685dd9315f0f08416168356d
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
nvme_rdma_req_get() is an internal function, and its only caller already
checks for a valid rqpair, so the NULL check is unnecessary.
Also clean up the redundant STAILQ_EMPTY/STAILQ_FIRST logic and use
STAILQ_REMOVE_HEAD.
Change-Id: Ic3828e8b5e881879173cb59350e39c5fac90e6ef
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
nvme_rdma_pre_copy_mem() does not have any failure cases, so remove its
return value and remove the never-taken branch in its only caller,
nvme_rdma_qpair_submit_request().
Change-Id: I91011734ed0c20f8db691d62172fe1a3021dd3a1
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
nvme_rdma_req_put() is an internal nvme_rdma.c function, and all of the
callers already have the rqpair, so pass it directly. We also already
verify that all of the callers have a valid rqpair and req before
calling nvme_rdma_req_put(), so it doesn't need to check for NULL
pointers.
This also means that spdk_nvme_rdma_req doesn't need to hold a pointer
to its rqpair anymore.
Change-Id: I893a46a9074f0a843e379d10c123f9292eb3b1a4
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The only place where outstanding_reqs was checked was in
nvme_rdma_req_put(), but the error case there could only happen if some
kind of internal programming error occurred (e.g. calling
nvme_rdma_req_put() on an invalid request).
Change-Id: I71e40ce562a8720dfaf70437ffd4c6493327c091
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
nvme_rdma_ibv_send_wr_init() was only called in one place, so just move
its contents into nvme_rdma_qpair_submit_request() since it allows
simplification of the code:
- req was always NULL, so remove the code that used req entirely.
- wr and sg_list are never NULL, so remove the checks for those.
Change-Id: I12a4f3502219d3681607686945e343f6808c0d2f
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
We currently don't handle discovery service referrals, so skip those, as
well as any other unknown subsystem type.
Change-Id: I64f889e9272fb57b5cf9bb5467b3abca3955baf5
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
QEMU's virtual NVMe controller device does not support the AER Set
Feature, so ignore its failure and continue.
Change-Id: I8b5c217a3112edabb6f76ec3e5f4ef774981a1d7
Signed-off-by: Cunyin Chang <cunyin.chang@intel.com>
Catch SIGBUS and handle it by remapping new memory into the
location where the BAR previously was.
Change-Id: Ie8d00a60a0bbe7f7ec57a5c39c0a63c5d9443206
Signed-off-by: Cunyin Chang <cunyin.chang@intel.com>
These functions will attach or detach from a PCI device. Attaching
typically means mapping the BAR.
Change-Id: Iaaf59010b8a0366d32ec80bb90c1c277ada7cfe7
Signed-off-by: Cunyin Chang <cunyin.chang@intel.com>
spdk_nvme_probe frees ctrlr when nvme_ctrlr_process_init is failed. But
ctrlr has already been freed while calling nvme_ctrlr_destruct. So
spdk_nvme_probe doen't need to free ctrlr.
The generic NVMe library controller initialization process already
handles enabling the controller; the RDMA transport should not need to
set EN itself.
For now, the discovery controller is cheating and not using the normal
initialization process, so move the EN = 1 hack to the discovery
controller bringup until it is overhauled to use the full
nvme_ctrlr_process_init() path.
The previous code where CC.EN was set to 1 before going through the
controller init process would cause an EN = 1 to EN = 0 transition,
which triggers a controller level reset.
This change stops us from causing a reset during the controller
startup sequence, which is defined by the NVMe over Fabrics spec as
terminating the host/controller association (breaking the connection).
Our NVMe over Fabrics target does not yet implement this correctly, but
we should still do the right thing in preparation for a full reset
implementation.
This patch also reverts the NVMe over Fabrics target reset
handling hack that was added as part of the NVMe over Fabrics host
commit to its previous state of just printing an error message.
Change-Id: I0aedd73dfd2dd1168e7b13b79575cc387737d4f0
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Most of the NOTICE level messages should have been TRACE.
Change-Id: Icbc4d398ab2580cf3a2349be11441b7a09603020
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Verify that qpair is not NULL before doing pointer math on it.
The NULL check after calling nvme_rdma_qpair(qpair) would not
trigger if qpair was NULL.
Fixes a crash if the Connect command failed, causing
nvme_rdma_ctrlr_create_qpair() to return NULL.
Change-Id: I158a5b1752892a7d5a72a9ac20c0c5b2cd781a81
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
The status.done flag polled by nvme_ctrlr_set_keep_alive_timeout()
was never initialized.
Change-Id: I323fae5f4ce12209a9699965ce07894bc3c6205a
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Use the NVMe over Fabrics spec definitions for TRTYPE rather than the
internal library transport type.
Change-Id: Idead559a8f8d95274fc580d10e82033822e6eda8
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
These need to be available for the lifetime of the probe_info structure,
so they can't be pointing at e.g. temporary buffers on the stack.
Change-Id: I5aaa898acf9314aab51600dd756f966965d37fd0
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
It always points to the same internal RDMA request complete function, so
just call that function directly.
Change-Id: Ic1fb6236bf43eaad62413df77d43be9ab855e5c7
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
We can't transfer more than the bounce buffer in a single command, so
report that rather than some bogus value.
Change-Id: I39b147916dcc2ee478470917298763a239a6a35a
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Record the user-provided asynchronous event configuration set via Set
Features, and return it in Get Features.
This value is not actually used, since AER is not implemented yet in the
virtual controller model, but it at least implements the mandatory
Set/Get Features.
This allows the hack in the NVMe host code that ignored the Set Features
failure to be reverted.
Change-Id: I2ac639eb8b069ef8e87230a21fa77225f32aedde
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Fill in the cached copy of CAP in the generic NVMe controller to match
the PCIe transport.
This is not really early enough, since CAP is used during the reset
process to determine the reset timeout, but that will have to be fixed
separately by rearranging some of the transport callbacks.
Change-Id: Ia8e20dbb8f21c2871afb9e00db56d0730e597331
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Make sure the entire NQN field is zero-padded, rather than using
strlen() on the input.
Change-Id: Icee68bd033feed057813beeb30cec102ed90840e
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
This fixes a compiler warning about unhandled enum cases in a switch.
Change-Id: Icecb56b47a05c13f390f03b877f8eae243b481a6
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
- add SPDK_NVME_OPC_KEEP_ALIVE to admin_opcode
- add SPDK_NVME_SC_INVALID_SGL_OFFSET, SPDK_NVME_SC_INVALID_SGL_OFFSET,
SPDK_NVME_SC_HOSTID_INCONSISTENT_FORMAT, SPDK_NVME_SC_KEEP_ALIVE_EXPIRED
and SPDK_NVME_SC_KEEP_ALIVE_INVALID to generic_status
The SPDK_TRACELOG macro depends on a CONFIG setting (DEBUG), so it
should not be part of the public API.
Create a new include/spdk_internal directory for headers that should
only be used within SPDK, not exported for public use.
Change-Id: I39b90ce57da3270e735ba32210c4b3a3468c460b
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Considering the process can be terminated in the cases like ctrl+c,
kill command or memory fault, the ref is tracked in the per process
structure spdk_nvme_controller_process and whenever there is other
process attaches or detaches the controller, a scan will be issued
to cleanup those unexpectedly exited processes.
Change-Id: Ib4f974f567a865748d42da4ead49edd383dfc752
Signed-off-by: GangCao <gang.cao@intel.com>
Instead of the next_sge callback returning the physical address
directly, make it return the virtual address and convert to physical
address inside the NVMe library.
This is necessary for NVMe over Fabrics host support, since the RDMA
userspace API requires virtual addresses rather than physical addresses.
It is also more consistent with the normal non-SGL NVMe functions that
already take virtual addresses.
Change-Id: I79a7af64ead987535f6bf3057b2b22aef3171c5b
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Function pointers will not work for the DPDK multi-process model (they
can have different addresses in different processes), so define a
transport enum and dispatch functions that switch on the transport type
instead.
Change-Id: Ic16866786eba5e523ce533e56e7a5c92672eb2a5
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
