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>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2244 (master)
(cherry picked from commit d3cf561199)
Change-Id: I75bf86e93674b4822d8204df3fb99458dec61e9c
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2510
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This is to backport below change to SPDK v20.01.2 LTS release.
6a29c6a906
Change-Id: I9b7ed97f2a376af71578ccb5556231832863b255
Signed-off-by: Liang Yan <liang.z.yan@intel.com>
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2262
Tested-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Solve the problem that the /sys/block/vd../max_segments is always 1 in the virtual
machine,and avoid the problem of low sequential read and write performance caused
by this limitation in the general block device layer of some lower kernels.
Backport this fix (9c6d4649eb)
to the SPDK LTS 20.01.2 release.
Change-Id: I30f6201bbfbb7885379b1b0ae19b64a1673e487f
Signed-off-by: suhua <suhua1@kingsoft.com>
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2261
Tested-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Xiaodong Liu <xiaodong.liu@intel.com>
iSCSI target got segmentation fault if connection is being exited
between spdk_iscsi_conn_write_pdu() and its callback
iscsi_conn_login_pdu_success_complete() are executed.
This was caused by recent asynchronous socket write feature.
Fixes issue #1278.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1275 (master)
(cherry picked from commit 628dc9c162)
Change-Id: Idffd90cd6ee8e6cb4298fe3f1363d8d5c5a3c49d
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1355
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: Ben Walker <benjamin.walker@intel.com>
Default was 8 which meant max of 8 bdevs. Bump it up to 64.
Fixes issue #1232
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1063 (master)
(cherry picked from commit 302f7aa6e4)
Change-Id: I966e90de5c27910df0e4da0d1062d9d1665f8de6
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1306
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The adminq poller could get a failure if the ctrlr has
already been hot removed, which starts a reset.
But while the for_each_channel is running for the reset,
the hotplug poller could run and start the destruct
process. If the ctrlr is deleted before the for_each_channel
completes, we will try to call spdk_nvme_ctrlr_reset() on
a deleted controller.
While here, also add a check to skip the reset if the
controller is already in the process of being removed.
Fixes#1273.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1253 (master)
(cherry picked from commit ba7b55de87)
Change-Id: I20286814d904b8d5a9c5209bbb53663683a4e6b0
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1305
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This isn't in the performance path, so using the mutex
here makes it a bit more consistent with other ctrlr
members such as 'destruct'.
This prepares for a future patch which will defer
ctrlr destruction on removal if a reset is in progress.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1252 (master)
(cherry picked from commit 2571cbd807)
Change-Id: Ica019cd90dc3b46ef6a13dd311054dbdc95855aa
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1304
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
We use `spdk_map_bar_rte()` to read mapped addresses
from PCI BARs.
This function is currently checking for NULL in each pair.
But in PCI memory, some registers can be left unused,
in which case they are set to 0.
As a result, we may read some NULL pointers from BARs,
which is OK.
To check if given address is indeed invalid, we should first
check if it is used.
So it is best to delegate such checks to the
user of this function.
In fact, users already do the NULL check where it is needed
(ex: virtio_pci.c:390, nvme_pcie.c:589)
so this patch just removes them from `spdk_map_bar_rte()`.
This solves github issue #1206
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1129 (master)
(cherry picked from commit d4653a31e0)
Change-Id: I88021ceca1b9e9d503b224f790819999cd16da01
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1302
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Fix Segmentation fault on the target side.
Issue:
rdma.c:2752:spdk_nvmf_rdma_listen: *NOTICE*: *** NVMe/RDMA Target Listening on 192.168.35.11 port 4420 ***
rdma.c: 789:nvmf_rdma_resources_create: *ERROR*: Unable to allocate sufficient memory for RDMA queue.
rdma.c:3385:spdk_nvmf_rdma_poll_group_create: *ERROR*: Unable to allocate resources for shared receive queue.
Segmentation fault (core dumped)
GDB:
Program terminated with signal 11, Segmentation fault.
736 if (resources->cmds_mr) {
(gdb) bt
736 if (resources->cmds_mr) {
(gdb) bt
0 nvmf_rdma_resources_destroy (resources=0x0) at rdma.c:736
1 0x0000000000497516 in spdk_nvmf_rdma_poll_group_destroy (group=group@entry=0x2fe1300) at rdma.c:3489
2 0x00000000004978bb in spdk_nvmf_rdma_poll_group_create (transport=0x2fe11d0) at rdma.c:3371
3 0x000000000048df70 in spdk_nvmf_transport_poll_group_create (transport=0x2fe11d0) at transport.c:267
4 0x000000000048a450 in spdk_nvmf_poll_group_add_transport (group=0x2f49af0, transport=<optimized out>) at nvmf.c:941
5 0x000000000048a6cb in spdk_nvmf_tgt_create_poll_group (io_device=0x2fce600, ctx_buf=0x2f49af0) at nvmf.c:122
6 0x00000000004a0492 in spdk_get_io_channel (io_device=0x2fce600) at thread.c:1324
7 0x000000000048a0e9 in spdk_nvmf_poll_group_create (tgt=<optimized out>) at nvmf.c:723
8 0x000000000047f230 in nvmf_tgt_create_poll_group (ctx=<optimized out>) at nvmf_tgt.c:356
9 0x000000000049f92b in spdk_on_thread (ctx=0x2f81b20) at thread.c:1065
10 0x000000000049f17d in _spdk_msg_queue_run_batch (max_msgs=<optimized out>, thread=0x1e67e90) at thread.c:554
11 spdk_thread_poll (thread=thread@entry=0x1e67e90, max_msgs=max_msgs@entry=0, now=now@entry=947267017376702) at thread.c:623
12 0x000000000049af86 in _spdk_reactor_run (arg=0x1e678c0) at reactor.c:342
13 0x000000000049b3a9 in spdk_reactors_start () at reactor.c:448
14 0x0000000000499a00 in spdk_app_start (opts=opts@entry=0x7ffc2a5e0ce0, start_fn=start_fn@entry=0x40aa80 <nvmf_tgt_started>,
arg1=arg1@entry=0x0) at app.c:690
15 0x0000000000408237 in main (argc=5, argv=0x7ffc2a5e0e98) at nvmf_main.c:75
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1073 (master)
(cherry picked from commit 9d93c08234)
Change-Id: Id9bf081964d0cf3575757e80fc7582b80776d554
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1301
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: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
This is to indicate the ABI breakage in the bdev library. A function's
argument list was changed which breaks both backwards and forwards
compatibility.
Going forward, all backwards compatibility breaking changes should be
marked with a rev of the SO major version for that library. All forwards
compatibility breaking changes should be marked with a rev of the SO
minor version.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1066 (master)
(cherry picked from commit c5911f0224)
Change-Id: I35e45c102c5c6de3c684919a10e5116f8f2c375f
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1300
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: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Users may only set the transport type, but for the actual probe
process, the trstring field is mandatory, so set the trstring
based on transport type at first. Also remove unnecessary
spdk_nvme_trid_populate_transport() call from each transport
module.
Fix#1228.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1001 (master)
(cherry picked from commit 8d6f48fbf8)
Change-Id: I2378065945cf725df4b1997293a737c101969e69
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1299
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
trstring variable in spdk_nvme_trid_populate_transport is not
initialized, that can lead to snprintf() writes some garbage to
trid->trstring if the user passes SPDK_NVME_TRANSPORT_CUSTOM trtype
Add return statement and assert to CUSTOM/default switch
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483469 (master)
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
(cherry picked from commit 3424def90a)
Change-Id: I6c6c37f9aa74d61b346f7be27fb890c7a34e9229
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1318
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
In the autotest, when calling kill_stub() function, there is error log
like this: "Device 0000:83:00.0 is still attached at shutdown!", so it's
better to detach the controller when exit the stub process.
But after call spdk_nvme_detach() in the stub process, there is another issue:
1. NVMe stub running as the primary process, and it will send 4 AERs.
2. Using NVMe reset tool as the secondary process.
When doing NVMe reset from the secondary process, it will abort all the
outstanding requests, so for the 4 AERs from the primary process, the 4
requests will be added to the active_proc->active_reqs list.
When calling spdk_nvme_detach() to detach a controller, there is a
assertion in the nvme_ctrlr_free_processes() at last to check the
active requests list of this active process data structure.
We can add a check before destructing the controller to poll the
completion queue, so that the active requests list can be flushed.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/977 (master)
(cherry picked from commit bad2c8e86c)
Change-Id: I0c473e935333a28d16f4c9fb443341fc47c5c24f
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1298
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Given enum was not aligned with spec. This status can be reported when
size equals 0.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/928 (master)
(cherry picked from commit a7a0d02d8b)
Change-Id: If51f6b051c13880c1fd4e6bb0a02f134b28b5a88
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1297
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: Jacek Kalwas <jacek.kalwas@intel.com>
Destroy subystem if spdk_nvmf_subsystem_set_sn or spdk_nvmf_subsystem_set_mn
failed. Check status in spdk_rpc_nvmf_subsystem_started callback, destroy
subsystem and report an error on error.
Fixes#1192
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/832 (master)
(cherry picked from commit c29247e1fe)
Change-Id: Id6bdfe4705b5f4677118f94e04652c2457a3fdcc
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1296
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This event can occur for either qpair or listening device. The
current implementation assumes that every event refers to a qpair
which is wrong. Fix: check if the event refers to a device and
disconnect all qpairs associated with the device and stop all
listeners.
Update spdk_nvmf_process_cm_event - break iteration if
rdma_get_cm_event returns a nonzero value to reduce the
indentation depth
Fixes#1184
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/574 (master)
(cherry picked from commit 804b066929)
Change-Id: I8c4244d030109ab33223057513674af69dcf2be2
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1295
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
On Fedora 30 we have noticed VPP 19.04 related issues:
1) Error values returned by vppctl in non-interactive mode
are not relevant to the success/fail of command.
Vppctl ALWAYS returns 0, so "-e" bash option is unable
to detect any errors.
2) We have intermittent pipefail errors (error 141) returned
by vppctl on disconnect from vpp, even though commands are
executed succesfully.
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Signed-off-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1214 (master)
(cherry picked from commit 7b7e97604b)
Change-Id: Ie22ea24f7e81017089b899111724d338eeb81113
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1287
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
It is possible for multiple blob persists to affect one another.
Either by blob->state changes or blob mutable data.
Safe way to prevent that is to queue up the persists.
Next persist will be executed only after previous one completes.
Fixes#1170Fixes#960
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/776 (master)
(cherry picked from commit 030be573f3)
Change-Id: Iaf95d9238510100b629050bc0d5c2c96c982a60c
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1308
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>
_spdk_blob_persist_check_dirty() function will be
called in subsequent patch at the end of persist
in _spdk_blob_persist_complete() to proceed
with any queued up persists.
Please see following patch for this.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/872 (master)
(cherry picked from commit dd80edb2b4)
Change-Id: Ieeb334e23cde329743647f728e70dd60333c224a
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1307
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>
On newer versions of FIO, there is an issue with heavy verify workloads
where one of the headers (rand_seed) gets incorrectly generated by fio
during verify. This can be circumvented by using the verify_backlog
flag.
This is needed because it will enable testing this workload on the tcp
transport using fio in the SPDK test pool.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/988 (master)
(cherry picked from commit 2b43f6353f)
Change-Id: I028be3fdb72a76733b4226a37b6332cd45d0f774
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1294
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
GCC9 complains:
./db/version_edit.h:134:71: error: implicitly-declared "constexpr
rocksdb::FileDescriptor::FileDescriptor(const
rocksdb::FileDescriptor&)" is deprecated [-Werror=deprecated-copy]
From what I see this can be fixed by explicitly
defining some constructors and assignment operators,
even setting them to `= default;`. I didn't dig into
this further, just ignore the warning for now.
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1082 (master)
(cherry picked from commit a5bcbbefcb)
Change-Id: Ia0ee0cc5fc1dce36f7098959d383b08855a825df
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1286
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Purpose: The function spdk_sock_request_put may
return an error code, and close the socket, so we should change the
return type of _sock_check_zcopy.
If the return value of _sock_check_zcopy is not zero,
we should not handle the EPOLLIN event.
Fixes#1169
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483311 (master)
(cherry picked from commit 9587017902becf2e1ce9a13dea1cd9ee4659a538)
Change-Id: Ie6fbd7ebff54749da8fa48836cc631eea09c4ab8
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483411
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
With recent changes to extent on-disk metadata format,
new format (Extent Pages) is not backwards compatible.
Meanwhile old format (Extent RLE) is backwards
compatible with older SPDK applications.
Summing up:
Blobstore created pre SPDK 20.01 can only use Extent RLE.
Blobstore created starting with SPDK 20.01 can use both,
Extent Pages and Extent RLE specified by use_extent_table opts.
When use_extent_table is set to true, invalid flag for it is set.
SPDK application pre 20.01, will not load such blob.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483220 (master)
(cherry picked from commit dc24539c40f28f0c85ca80b823ba3f17d005b172)
Change-Id: If14ebd03f19eb581d71dcb46191e099336655189
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483395
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Github issue 1165 details some issues we have with soft-roce and these
tests. Right now we are disabling them for build stability.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483300 (master)
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
(cherry picked from commit e0f63b969d)
Change-Id: I3a9e28ff3cc1c6ac7d9aa91d93541e295514bb7b
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483407
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This allows us to avoid trying to map the same physical address to the
IOMMU in physical mode while still making sure that we don't
accidentally unmap that physical address before we are done referencing
it.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483133 (master)
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
(cherry picked from commit f4a63bb8b3)
Change-Id: I947408411538b921bdc5a89ce8d5e40fd826e971
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483256
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
The fuse command value is a two byte value, but we were only checking to
see if the fuse value was equal to SPDK_NVME_CMD_FUSE_FIRST or
SPDK_NVME_CMD_FUSE_SECOND in spdk_nvmf_ctrlr_process_io_fused_cmd. If a
haywire initiator sent a command with a fused value equal to
SPDK_NVME_CMD_FUSE_MASK, that would result in us skipping all checks and
dereferencing a null pointer in
spdk_nvmf_bdev_ctrlr_compare_and_write_cmd.
To fix this, add an extra condition to validate the cuse field.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483123 (master)
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
(cherry picked from commit f0ca01e102)
Change-Id: I1ec4169ff5637562effd694f7046c6e3389627f1
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483255
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
When structure for output of json decoders in not initialized
spdk_json_decode_string may fail trying to free uninitialized
string.
This patch changes mallocs used to allocate context and structure
for output of decoder with calloc.
Fixes#1151
Change-Id: I180b2ec52350b4ca90e7c318b4f2d13af554ec49
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/483107
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
This was observed after running nighly tests on previous patch.
As part of it, autopackage.sh compiles SPDK
without debug flag set. Exposing the uninitialized var here.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Iedb1641f3c0d4a21f293c81cd4fcf35c6d1c7ae5
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482893
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Extent table and extent page descriptors are now
set to be default way clusters are serialized on disk.
With this patch UT are ran with and without
extent table.
Changed two asserts in test, since amount is dependent on
which type of serialization is used.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ica58fce6a4effd014d7dd40ee26edd0fa3196d0f
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/481901
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
This will be used to add another run of whole UT suite
with extent pages on/off.
Next patch in series will be enabling both types of
extent serialization for all UT.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ia8b4b8822edefb90ffc13cf777885f9af95e4545
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482170
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
All the non-ext version of the call is doing, is calling
ext with NULL as opts. Then default opts are used in
its place.
This change was facilitated by next on in series,
where all blob opts will be initalized in UT
with parameter use_extent_table either set to
true or false.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I62b642c1808b38a5f7c94a5900f25f4978a4ec39
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482859
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
ctx->extent_page signifies if page was allocated
for insertion.
1) It is possible for a thread to claim extent page
on its own thread, and put it in ctx->extent_page.
If conflicting thread allocates another ctx->extent_page,
then it should be freed. This does not mean failure
to insert cluster. As different threads could have
been trying to allocate different clusters,
so condition on line 6716 does not cover it.
If so then it shouldn't be an issue to release
the claimed ctx->extent_page and proceed with updating the
extent page which originally won the race.
NOTE: if clusters were conflicting, then extent_page is
freed in _spdk_blob_insert_cluster_cpl().
2) At this point of _spdk_blob_insert_cluster_msg()
we already verified that there already is
extent page allocated at "*extent_page".
In such case ctx->extent_page will be 0,
and should not be used.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Id5b57c88248890eee60d2e7dbecbd984c98b561b
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482867
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
sz is set to number of clusters that should be have been
in particular unallocated EP (remaining_clusters_in_et
up to SPDK_EXTENTS_PER_EP).
The cluster array should be set to 0 only in region
between original size (cluster_array_size) and new
total size (active.num_clusters).
It was incorrectly using sz in the memset.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ic43e89c17d53e9529e3ed0349aeb4fb7dc6593f2
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482858
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Previously part of function assumed that cluster count
1)means number of clusters in EP and another 2) that it is
following the active.num_clusters (akin to extent_rle).
This was incosistent and showed when using multiple
extent pages to serialize metadata.
This patch changes it to only go with 1), so it is clear
that it means number clusters within particular EP.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I155104cabc127ed47df04434032fb01e08948e13
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482848
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Following changes are done in this patch:
1) _spdk_blob_serialize_extent_table now persists
at least single extent page. When num_extent_pages == 0.
2) Minimum valid size of ET descriptor is even without
extent_pages. This is a case when there are no EP,
but we still want to persist num_clusters in ET.
3) Taking above points, redone the loop for serializing
extent pages.
4) Make sure to mark blob dirty if any new extent pages
were allocted.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I81dc6cf2de2722bb49927ed42f4b9f31292f78c5
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482847
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Changed assert checking if cluster array is allocated
when loading extent pages. This is true only for
the first extent page being loaded, of course after that
the cluster array can be already allocated.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I0f96294ede5a12ffd6bca73cbeadba8d94a35bac
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482857
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Size of a blob (thus size of clusters array in mutable data)
is known from extent table descriptor.
Extent pages were read sequentially in order they were
placed in extent table. This meant that cluster
array could have been filled up from beginning to end.
Yet reading extent pages in any other order,
would result in incorrect placement of clusters.
This patch adds first cluster index that is contained within
each extent page. This will allow to read/write
multiple extent pages in parallel, since
we will know where in clusters array to put the cluster idxs.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ib6b9332111cd93f990d057dc60624152907dd87f
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/482701
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: Jim Harris <james.r.harris@intel.com>
