In order to fit inside the maximum memory usage limit, recovery needs
to be split into multiple parts. During each iteration, part of L2P needs
to be read, modified as necessary and saved back to the cache.
This patch introduces the load/save steps, initialization of seq_id array
and valid map recovery. The actual L2P recovery is done in the followup patch.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I8ceadc5ef280542a173d83b932a983d5d86604a1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13371
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Adds recovery of open bands from P2L metadata region. Recovers the
commited P2Ls and write pointers for them.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I943c53f55e653dd075035cef7ddba448c990be87
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13370
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Recovers the open/close/free state of bands after shutdown, initializing
necessary lists.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I4a6bd4ed1013ce8d04f44d1772dcd1f0e4e365bd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13368
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Since base device doesn't require VSS, FTL introduces a mechanism that
will allow for recovering both the P2L and write pointer of open bands
after a dirty shutdown. After writing 1MiB of data to a band, a 4KiB
block describing the P2L will be persisted to cache device, effectively
emulating VSS for the base device.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: Ic6be52dc09b237297a5cda3e752d6c038e98b70e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13367
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Track the relative sequence of opening and closing bands and chunks.
Necessary for detecting the most recent user data during dirty shutdown recovery.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I682030e58284d7b090667e4e5a9f4bbc7615708a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13366
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Adds ability for FTL to startup after clean shutdown.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I2f1b83bb3eb1487b6665c95e76c48881e8899b16
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13364
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>
Adds additional debugging functionality - ability to check the validity
of all L2P entries and valid map to check for inconsistencies after FTL
startup. Since this is a very time consuming process, it's controlled by
an environment variable and not executed during normal operations.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I4766a1576c058f69fa047f45d2d8be6d0ad0b3cd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13363
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>
Adds necessary functions for setting up the state of FTL components
based on loaded in metadata.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I3a4c05230c877850e61d4f31d495d38121d27b3f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13362
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Adds the ability to persist only the most important metadata. The rest
is stored in shared memory.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I4084c04ba09115a7a08ff66fd33552a2ec60d801
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13360
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Add an extra step during FTL shutdown to save all metadata.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: Idc2f77e15bbd02028548cc88355cd450175830e8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13359
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Flat L2P (all L2P in memory) needs to be specifically built against, due
to large memory consumption for big devices.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: Ib8906e10868455f88725b69b2b033b70a9f7256c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13358
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Adds eviction of least recently used pages from the L2P cache - dirty
pages will be persisted.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: Ic646f7e9da777d077b5cb9b409c3f03ef05b1273
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13357
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Adds paging in from the cache device to memory.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I250009d12e9ed5ad52ee861ec5157cf983cf8cfc
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13356
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
There is a set amount of pinned pages available. If exceeded they will be deferred
and processed in the future, using eviction logic.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: Ic642a5870db009ccf57152dd8a4178a6b2098ee1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13355
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This commit also introduces ranking pages, based on usage for determining the least used
page to be evicted.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: Iaf3812177b61376bb38aa209e4ba8576d784ffb5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13354
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
L2P cache allows for partial storing of L2P in memory, paging in and out
as necessary, lowering the total memory consumption.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I727fec9d2f0ade4ca73e872d62a2ec10cfdb0a88
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13353
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Adds a debug function, that scans the whole P2L of band, when
it's getting closed. The P2L is compared against both L2P and
valid map to check for any discrepancies.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: Ia4d7be65415e6af3752d676de69b6fdcb73effb4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13352
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>
Utilize the valid map when picking physical blocks to compact/relocate,
speeding up the process.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I860e3cf25a5907591e4f3043def67156fec8b0df
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13351
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Adds P2L validity map tracking - a bitmap marking all physical LBAs
as containing valid (current) user data or not. A clear bit denotes
the location has no valid data and may be skipped during relocation
or compaction. A set bit means it may have valid data (it's still
necessary to do the necessary comparision against L2P).
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I6a831a97b3080eb7c880d9c4feab41b523467885
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13350
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>
Adding API for the bringup part of fast shutdown/startup. Adds
shared memory utilization for necessary functions during initialization.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: Iab2da102fd0ccaa56fbdb9b3c765be5eeefff145
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13349
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Adds API for fast shutdown - the ability for FTL to skip most
of the metadata persists made during clean shutdown, and relying
on their representation in shared memory instead. This allows for
faster update of SPDK (or just FTL, assuming no metadata changes),
with downtime reduction from 2-5 seconds to 500-1000 ms (for
14TiB+800GiB base and cache drives).
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I5999d31698a81512db8d5893eabee7b505c80d06
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13348
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>
Helper functions which determine which md regions will be stored
in shm.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I94cbfca66dfb56457a350874dbd1de63a2e07661
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14159
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>
Stores P2L map of open bands in shared memory, allowing for
faster recovery times from application crash.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I519441af05e4d0f57768835bf01c800556873c58
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13347
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>
Allows for using shared memory in memory pools. Adds API for
accessing such pools after dirty shutdown (claiming them, ie.
marking an entry as actively used; calling the
ftl_mempool_initialize_ext will reclaim all unused entries back
to the pool). Also introduces API for accessing objects, since
using direct pointers is not possible (as addresses may change
inbetween application startups).
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I5325b39d68aef7e231945cee9d92c925cab2fb2a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13346
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Main use case is to allow for keeping it in shared memory, to
speed up the recovery time after application crash.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I36b6b8331cd6483c5bd202e5f9103c351d705da8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13345
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>
Relocation will
1. Read LBA map of a given band
2. Pin the LBAs
3. Issue writes of valid LBAsto the new location
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: Ie753a790e56a86bfa1e451b5eda78b88eeacd3cb
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13344
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Adds functions for reading end metadata and initializing band reloc state.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I3d12c4a7edd36f0437bf10316114c83efe449f0f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13343
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>
After chunk is compacted it can be moved to the free state, able to be used
for new user IO again.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I7f9c341169b171ee246c5aa161d74903b91bdc2f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13338
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>
During compaction FTL moves valid user data from the nv cache drive to
the bottom device.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: Ia200af39cec80014fac3a10f20d2859b10a81088
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13337
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>
Add writer - tracks and manages band state transitions and write pointer as IO
is issued to it.
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I5f878dc15bc1c1ac84835f75fe440672fad541d5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13335
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>
Calculates general priorities and trigger points for writers
(gc and compaction) dependent on number of free bands. GC will
be started at SPDK_FTL_LIMIT_START level, while at SPDK_FTL_LIMIT_CRIT
compaction needs to be stopped and only GC is allowed to work.
This is done to make sure FTL doesn't run out of free bands and deadlock
itself.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Change-Id: I1aab98503c2e79e97f8e4e9fb1257530fa9770e2
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13326
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>
This is targeted to fix the following error seen under clang:
ftl_nv_cache.c:54:1: error: unused function 'nvc_data_blocks'
[-Werror,-Wunused-function]
nvc_data_blocks(struct ftl_nv_cache *nv_cache)
^
Signed-off-by: Michal Berger <michal.berger@intel.com>
Change-Id: I11d52e76df5872819770d9468b6fa4ae54d8927c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14055
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: <sebastian.brzezinka@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
