VMD section with Enable flag set to true
need to be defined in config file to
enumerate devices behind VMD.
Change-Id: I0b35d93b224025050ae0c081af720ed816c9f0fa
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459765
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Return immediatly from ftl_reloc_add() if no
new blocks was added to relocation.
Change-Id: If80dfa725e0bb9f3b8987740012858a671c5ad90
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457626
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Added check if band that is added to reloc
have any valid blocks. Return immediately
if there is no valid blocks.
Change-Id: I2bce088e0ad71479c6899fff96845397d12e2e92
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457625
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In case ANM event occurs on band being
relocating (band is on active reloc queue)
we shoudn't add such band to pending queue.
Change-Id: I92a8bee11309097e19afaea549460f1d4387e3e5
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458617
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In case high priority band was added
for relocation it should be removed
from active/pending queue if it was
already on one of them.
Change-Id: Id0591b1d3a4174dd05eb1c32227e4d3b3a9cbcd0
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458057
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
In case ANM event occurs on open band there
can be situation that reloc will try to read
block on which there is ongoing write.
This is happening because lba valid map is
updated before write submission to allow sent
consistent metadata to disk before all user
writes are completed.
Added write offset to the each chunk and
add check to reloc if particular ppa is written
on that chunk.
Change-Id: Ic95a06e69381d2152a86984b65a0975afaff955d
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458056
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In case ANM event occurs on open band
reloc need to be able to process such
event.
If band is not in closed sate do not
alloc lba map for it and do not set
it to free state after relocation.
Change-Id: I2f4a5770fef08271d222936ca19f3cc98e5e5be1
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457612
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Open bands need to have lba map segments
set to cached state to prevent read lba
map from disk during relocation events.
Change-Id: Ib4f1ed19131fad174c1d2f70e4c02e83701e2a0a
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457853
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Band tail PPA should be initialized when
new FTL instance is created.
Change-Id: Ie2fb72aa3f29eece0b6f8912998b33af3ba6b355
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457777
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Send ANM event to core thread for further
processing. This will remove a need of locking
in relocate module when ANM event occur.
Change-Id: I0efb1f1b8c96c107cda5fe78e8ee5672cde39f11
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457611
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
In our previous code, we will handle all the PDU until there is
no incoming data from the network if we can continue the loop.
However this is not quite fair when we handling multiple connections
in a polling group.
And this change is setting a maximal NVME/TCP PDU we can handle
for each conneciton, it can improve the performance. After some
tuing, 32 should be a good loop number. Our iSCSI target uses
16.
The following shows some performance data:
Configuration:
1 Command used in the initiator side:
./examples/nvme/perf/perf -r 'trtype:TCP adrfam:IPv4 traddr:192.168.4.11 trsvcid:4420'
-q 128 -o 4096 -w randrw -M 50 -t 10
2 target side, export 4 malloc bdev in a same subsystem
Result:
Before patch:
Starting thread on core 0
========================================================
Latency(us)
Device Information : IOPS MiB/s Average min max
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 51554.20 201.38 2483.07 462.31 4158.45
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 51533.00 201.30 2484.12 508.06 4464.07
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 51630.20 201.68 2479.30 481.19 4120.83
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 51700.70 201.96 2475.85 442.61 4018.67
========================================================
Total : 206418.10 806.32 2480.58 442.61 4464.07
After patch:
Starting thread on core 0
========================================================
Latency(us)
Device Information : IOPS MiB/s Average min max
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 57445.30 224.40 2228.46 450.03 4231.23
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 57529.50 224.72 2225.17 676.07 4251.76
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 57524.80 224.71 2225.29 627.08 4193.28
TCP (addr:192.168.4.11 subnqn:nqn.2016-06.io.spdk:cnode1) from core 0: 57476.50 224.52 2227.17 663.14 4205.12
========================================================
Total : 229976.10 898.34 2226.52 450.03 4251.76
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I86b7af1b669169eee2225de2d28c2cc313e7d905
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459572
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: Ben Walker <benjamin.walker@intel.com>
SPDK blobfs has asynchronous APIs defined in blobfs_internal.h file,
as users may want to use them, so we remove them to the public .h file.
Change-Id: I1835d97060101f6315a73cb8638b15ff7e13ba54
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457547
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
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>
I tried experimental binding of SPDK with Mongo-rocks. However,
the binding sometimes invoke blobfs APIs without thread
initializations. In that case, null dereferences occur. In other
words, we need to carefully use blobfs not to invoke any threads
that are not registered to blobfs.
This patch simply adds a sanity check at every use of blobfs APIs.
By doing this, we do not need to care about which threads can use
blobfs APIs.
Change-Id: I5b37b0267306a7c76d20e81c1773a6a33be7828c
Signed-off-by: Takeshi Yoshimura <t.yoshimura8869@gmail.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/418966
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
"C2HSuccess" is only valid for TCP transport. So this parameter
should be looked up only for TCP transport. Without the change,
spdk_nvmf_parse_transport() would bailout early for RDMA and
other transports without every creating them.
Signed-off-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Change-Id: I34bdff2f4ab930516743cd5dbf022d75e60fd85c
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459571
Reviewed-by: Maciej Szwed <maciej.szwed@intel.com>
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: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In DIF library there are many functions that converts offset or
length from LBA based to extended LBA based.
Factor out them by adding a helper function _to_size_with_md().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Id5576edacc8a07095726f659c4b53ac3aa83727d
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459530
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This will be used to get extended LBA based range or length in NVMe/TCP
target later.
Change-Id: Id0f08bdaeea634dbc05b34a0f7914be21aef9aae
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458706
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Add spdk_dif_update_crc32c_stream to update CRC32C by stream fashion.
spdk_dif_update_crc32c_stream utilizes the updated _dif_update_crc32c_split.
A minor bug was found in UT for spdk_dif_update_crc32c and is fixed
together in this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I92358e845e8e2e17c6f288aa718b947e71e6e1fb
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458919
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
For NVMe/TCP target, data segments which correspond to H2C or C2H PDU
will have any alignment, and _dif_update_crc32c_split will have to
process partial data block, particularly the following types:
- start and end are both within a data block.
- start is within a data block, and end is at the end of a block
On the other hand, _dif_update_crc32c_split had assumed that passed
block is always a complete block.
This patch exposes offset_in_block, data_len, and guard as parameters
of _dif_update_crc32c_split() and make _dif_verify_split() process
the above two types of data block properly.
The next patch will utilize the updated _dif_update_crc32c_split to
add spdk_dif_update_crc32c_stream().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Iee29377ad49d4f209673fffb4de4a23a54f31766
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458918
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
spdk_dma_*malloc() is about to be deprecated.
Change-Id: I6c308ee546c28c479ceb903bc1749bf5209dc6fe
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448172
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: <uma.willpower@gmail.com>
spdk_dif_generate_stream() and spdk_dif_verify_stream() are very
similar. Factoring out the common part into a function will improve
the maintainability and do in this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I16ecd0860c75037d9182298d7513749dfe8e9b56
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458376
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Add spdk_dif_verify_stream to verify DIF by stream fashion.
spdk_dif_verify_stream utilizes the updated _dif_verify_split.
spdk_dif_verify_stream is very similar with spdk_dif_generate_stream().
UT code demonstrates how it is realized.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I1c5d197cf4c0bbc82c8e7f4fa45ddc0b94051058
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458330
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
For NVMe/TCP target, data segments which correspond to H2C or C2H PDU
will have any alignment, and _dif_verify_split will have to process
partial data block, particularly the following types:
- start and end are both within a data block.
- start is within a data block, and end is at the end of a block
On the other hand, _dif_verify_split had assumed that passed block
is always a complete block.
According to the refactoring done in the last patch, this patch
exposes offset_in_block, data_len, and guard as parameters of
_dif_verify_split() and make _dif_verify_split() process the
above two types of data block properly.
The next patch will utilize the updated _dif_verify_split to
add spdk_dif_verify_stream().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic371d3ccefbd5fe8147a948a624013be2702128e
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458329
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
For NVMe/TCP target, data segments which correspond to H2C or C2H PDU
will have any alignment, and _dif_verify_split will have to process
partial data block, particularly the following types:
- start and end are both within a data block.
- start is within a data block, and end is at the end of a block
On the other hand, _dif_verify_split had assumed that passed block
is always a complete block.
To process the above types, separating guard computation, DIF copy
and skipping metadata field, and DIF verification into three parts
will be helpful and is done in this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic4f1765e01507efa812dfaf7a8018666c6346f8e
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458328
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Some of the errors were silent, making it hard to pinpoint the exact
failing call. This patch adds SPDK_ERRLOGs for each error path.
Change-Id: I71be6c97cab916ac52314e5f4e4d63358877bd96
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458426
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Send LBA along with the data block when mirroring writes to the
non-volatile cache. The metadata buffer is retrieved from the metadata
pool, so the maximum number of concurrent requests is limited to
nv_cache.max_request_cnt, while the number of blocks in a single request
is limited by nv_cache.max_requets_size.
Change-Id: If260302d16039183fb0fe073ef7419947532cfab
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458093
Reviewed-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
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>
Initialize the memory pool for storing metadata (LBAs) when writing data
to the non-volatile cache. The mempool's object count and size can be
configured via nv_cache.max_request_cnt / nv_cache.max_request_size
respectively.
Change-Id: I376df9a75be13d4b29ba475f350edf402c868d48
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458092
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Address translation wasn't correct for >32 bit length packed address.
This commit fixes the issue and adds a corresponding unit test.
This patch fixes issue #774:
https://github.com/spdk/spdk/issues/774
Signed-off-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Change-Id: Idce67c47f2a9888f9e2ae2eadaf71ccc34e5c260
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457114
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Added spdk_pci_get_first_device() and
spdk_pci_get_next_device() to iterate
over all devices on g_pci_devices list.
Change-Id: I65079fb3e274195707dee64bc1fb8b4b72d07352
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450924
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
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>
Put the locks inside cleanup_pci_devices().
This serves as cleanup.
Change-Id: I040b28006e5584d1f33af26b63cafedbafe04fdb
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458934
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: John Kariuki <John.K.Kariuki@intel.com>
The global pci tailq is no longer modified on the dpdk
thread, so on the spdk thread we can access it safely
without any lock. The code is slightly more readable
then.
This shows that cleanup_pci_devices() is always wrapped
with lock/unlock. We'll put the locks inside this
function in the next patch.
Change-Id: Ia4d386b78a87078761df0a3b953bfc4ff44102f8
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458933
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>
To safely access the global pci device list on an spdk
thread, we'll need not to modify this list on any other
thread. When device gets hotplugged on a dpdk thread,
it will be now inserted into a new global tailq that
can be accessed only under g_pci_mutex. Then any
subsequently called public pci function will add it to
the regular device tailq.
Change-Id: I9cb9d6b24fd731641fd764d0da71bedab38824c9
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458932
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>
To safely access the global pci device list on an spdk
thread, we'll need not to modify this list on any other
thread. When device gets hotremoved on a dpdk thread,
it will now set a new per-device `removed` flag. Then
any subsequently called public pci function will remove
it from the list.
Change-Id: I0f16237617e0bea75b322ab402407780616424c3
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458931
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>
For VMD driver we'll need to introduce some way of
iterating over all spdk pci device objects and we would
like to achieve that with simple spdk_pci_get_first_dev()/get_next_dev()
APIs. To make it thread safe though, we would have to
expose some public pci mutex to be locked around the
iteration and we don't want to do that, so we'll make
PCI APIs usable from only a single thread - this will
prevent any pci devices from being removed inbetween
subsequent get_first/get_next calls.
We currently have the following players accessing pci
device state:
1) public APIs, obviously (on any thread right now)
2) VFIO hotremove callback (dpdk interrupt thread)
3) rte_eal_alarm for detaching rte_pci_devices (dpdk
interrupt thread)
4) DPDK hotplug IPC (dpdk interrupt thread)
There is g_pci_mutex providing the thread safety, but
even today it doesn't protect #3 and #4, making the
entire pci layer prone to data corruption.
To make #3 and #4 safe, we would have to lock inside
device init/fini callbacks (spdk_pci_device_init/fini),
but those are called directly inside the public device
attach/detach functions which already lock.
So now, with the decision to drop thread safety from
public pci APIs, we narrow down the locks inside public
functions and introduce locks inside those lower-level
init/fini callbacks.
Change-Id: I5dcbc9cdcbab65ee76cd3c42890f596069ec9a8a
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458930
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>
Converting to the struct virtio_req is useless.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I141268314d28cf87bdef529808c8e18bd1b41c9d
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459360
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Currently we are missing cleanup routine for case when
power failure interrupts creating snapshot. This patch
add such routine.
For the case where we find blob with a parent snapshot ID
matching newly created snapshot we can finish whole process
during recovery by processing forward with setting snpashot
as read only, removing xattr and syncing. We should remove
snapshot only if there is no blob with parent pointing at
snapshot.
Fixes github issue #760
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: I2f0e298164e07a2b4dfa5367e8878facef640702
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455216
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
By default QAT will be selected if available however a new RPC
can be used to either auto-select (default) or specify either
ISAL or QAT.
Change-Id: I37cf7640bbd8cef455583e1eccb8adb59cc419d8
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456693
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: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Fixed issue when restoring from a dirty shutdown - sometimes end md
wasn't erased after a band was prepared from writing when a shutdown
happened. This resulted in inconsistency between the new head md and
old tail md, which was technically valid. Band sequence numbers would
then be reused, causing a failure on any subsequent restore.
Signed-off-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Change-Id: Ic3e968be02bb814d6c85f0a3279403fe99337b86
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459287
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Changed to use 4k alignment in dirty shutdown I/Os. Otherwise the
scatter gather lists used in QEMU for underlying file/block device
would use an extra entry (e.g. 17 for 16 sector writes), and eventually
some I/Os would write to offset 0 in underlying file, corrupting head
metadata.
Signed-off-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Change-Id: If8c88ce708529b094a09c8ee952912cc22cd53b9
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458090
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Band's lba_map needs to be set to NULL before restore completes, as
it's not allocated on a per band basis and instead uses a pool from
restore struct itself. Without the fix initializing a band for writing
would hit an assert during proper allocation in ftl_band_alloc_lba_map.
Signed-off-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Change-Id: Icff4f54cbe722cb6030b9dfd55726b9b0d6c1e27
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458422
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This patch makes spdk_dif_set_md_interleave_iovs() and
spdk_dif_generate_stream() process unaligned start of data segment
properly by using ctx->data_offset.
Separating this patch into two may be required but this patch is
small and aggregating into a patch is good to test.
UT code demonstrates how it is realized.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Idb5250aba4e12a34102e5ce067d725c685681177
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458142
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
To process unaligned data segment properly when a whole data buffer
is splitted into multiple data segments and each data segment has
any alignment, we have to update only data offset of DIF context
according to the progress.
Hence this patch adds an new API spdk_dif_ctx_set_data_offset().
The API will be used in the next patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I346ab583518b80792ea40d34cf0c8536ecc3d904
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458141
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
There're outstanding requests in spdk_jsonrpc_parse_request which caused by
connection close.
There are methods to call spdk_jsonrpc_server_conn_close, including
spdk_jsonrpc_server_conn_remove and spdk_jsonrpc_server_shutdown,
Some rpc methods call these functions to terminate connections ,that leads to
memory leaks.
Try to free outstanding requests after deciding to terminate a connection.
And do this follwing with close(conn->sockfd).
Fix issue #784, and can resolve other similar memory leaks about this.
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Change-Id: Icd287bd0c5670ee8ec32750b999f82b0fa89cf84
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458438
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
By now (5.1 is released), the Linux kernel initiator supports the
success optimization and further, the version that doesn't support
it (5.0) was EOL-ed. As such, lets open it up @ spdk by default.
Doing so provides a notable performance improvement: running perf with
iodepth of 64, randread, two threads and block size of 512 bytes for 60s
("-q 64 -w randread -o 512 -c 0x5000 -t 60") over the VMA socket acceleration
library and null backing store, we got 730K IOPS with the success
optimization vs 550K without it.
IOPS MiB/s Average min max
549274.10 268.20 232.99 93.23 3256354.96
728117.57 355.53 175.76 85.93 14632.16
To allow for interop with older kernel initiators, we added
a config knob under which the success optimization can be
enabled or disabled.
Change-Id: Ia4c79f607f82c3563523ae3e07a67eac95b56dbb
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457644
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: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
If users set the persist through power loss configuation file,
that means the Namespace has the capability to support ptpl
feature, here we added a ptpl_activated flag to indicate that
the users enable the feature or not. Users can use Set features
or Reservation Register commands to change the value.
Change-Id: Iae3fd44085c5be5bf9574e49efa567e8212dee20
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455906
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>
Vhost testing crashed from Nightly testing, because a member
access within null pointer of type 'struct ibv_send_wr'.
Change-Id: If8f34f23864883ea73516d2d1fe3b30137c04316
Signed-off-by: Hailiang Wang <hailiangx.e.wang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458913
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>
When JSON config was used, app layer was calling the
app start callback twice - once from internally-sent
"start_subsystem_init" RPC, and once from the app layer
itself.
In case of JSON configs, the callback from within the
RPC was actually called prematurely, as the real RPC
server was still starting in the background at that
point. We still need to start the app from that RPC in
case of `--wait-for-listen` option, but for JSON configs
it doesn't make sense. Just ignore it now and rely on
json config load completion callback to start the app.
Fixes#816
Change-Id: Ib54d624f3167137216c910b2d947bbd1dc5023b1
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458351
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
If reading the JSON config file has failed, we entered
spdk_app_json_config_load_done(-ERRNO) and tried to
close a client connection that was never initiated,
which resulted in NULL dereference.
To fix it, just check if client_conn != NULL before
attempting to close it.
Change-Id: I7340567c45e795f77110c2914e94ba83fa8d1bff
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458350
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
