Spdk/module/blob/bdev/blob_bdev.c

531 lines
15 KiB
C
Raw Normal View History

/* SPDX-License-Identifier: BSD-3-Clause
* Copyright (C) 2017 Intel Corporation.
* All rights reserved.
* Copyright (c) 2022-2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
*/
#include "spdk/stdinc.h"
#include "spdk/blob_bdev.h"
#include "spdk/blob.h"
#include "spdk/thread.h"
#include "spdk/log.h"
#include "spdk/endian.h"
#define __SPDK_BDEV_MODULE_ONLY
#include "spdk/bdev_module.h"
struct blob_bdev {
struct spdk_bs_dev bs_dev;
struct spdk_bdev *bdev;
struct spdk_bdev_desc *desc;
bool write;
int32_t refs;
struct spdk_spinlock lock;
};
struct blob_resubmit {
struct spdk_bdev_io_wait_entry bdev_io_wait;
enum spdk_bdev_io_type io_type;
struct spdk_bs_dev *dev;
struct spdk_io_channel *channel;
void *payload;
int iovcnt;
uint64_t lba;
uint64_t src_lba;
uint32_t lba_count;
struct spdk_bs_dev_cb_args *cb_args;
struct spdk_blob_ext_io_opts *ext_io_opts;
};
static void bdev_blob_resubmit(void *);
static inline struct spdk_bdev_desc *
__get_desc(struct spdk_bs_dev *dev)
{
return ((struct blob_bdev *)dev)->desc;
}
static inline struct spdk_bdev *
__get_bdev(struct spdk_bs_dev *dev)
{
return ((struct blob_bdev *)dev)->bdev;
}
static void
bdev_blob_io_complete(struct spdk_bdev_io *bdev_io, bool success, void *arg)
{
struct spdk_bs_dev_cb_args *cb_args = arg;
int bserrno;
if (success) {
bserrno = 0;
} else {
bserrno = -EIO;
}
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, bserrno);
spdk_bdev_free_io(bdev_io);
}
static void
bdev_blob_queue_io(struct spdk_bs_dev *dev, struct spdk_io_channel *channel, void *payload,
int iovcnt, uint64_t lba, uint64_t src_lba, uint32_t lba_count,
enum spdk_bdev_io_type io_type, struct spdk_bs_dev_cb_args *cb_args,
struct spdk_blob_ext_io_opts *ext_io_opts)
{
int rc;
struct spdk_bdev *bdev = __get_bdev(dev);
struct blob_resubmit *ctx;
ctx = calloc(1, sizeof(struct blob_resubmit));
if (ctx == NULL) {
SPDK_ERRLOG("Not enough memory to queue io\n");
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, -ENOMEM);
return;
}
ctx->io_type = io_type;
ctx->dev = dev;
ctx->channel = channel;
ctx->payload = payload;
ctx->iovcnt = iovcnt;
ctx->lba = lba;
ctx->src_lba = src_lba;
ctx->lba_count = lba_count;
ctx->cb_args = cb_args;
ctx->bdev_io_wait.bdev = bdev;
ctx->bdev_io_wait.cb_fn = bdev_blob_resubmit;
ctx->bdev_io_wait.cb_arg = ctx;
ctx->ext_io_opts = ext_io_opts;
rc = spdk_bdev_queue_io_wait(bdev, channel, &ctx->bdev_io_wait);
if (rc != 0) {
SPDK_ERRLOG("Queue io failed, rc=%d\n", rc);
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
free(ctx);
assert(false);
}
}
static void
bdev_blob_read(struct spdk_bs_dev *dev, struct spdk_io_channel *channel, void *payload,
uint64_t lba, uint32_t lba_count, struct spdk_bs_dev_cb_args *cb_args)
{
int rc;
rc = spdk_bdev_read_blocks(__get_desc(dev), channel, payload, lba,
lba_count, bdev_blob_io_complete, cb_args);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, payload, 0, lba, 0,
lba_count, SPDK_BDEV_IO_TYPE_READ, cb_args, NULL);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
static void
bdev_blob_write(struct spdk_bs_dev *dev, struct spdk_io_channel *channel, void *payload,
uint64_t lba, uint32_t lba_count, struct spdk_bs_dev_cb_args *cb_args)
{
int rc;
rc = spdk_bdev_write_blocks(__get_desc(dev), channel, payload, lba,
lba_count, bdev_blob_io_complete, cb_args);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, payload, 0, lba, 0,
lba_count, SPDK_BDEV_IO_TYPE_WRITE, cb_args, NULL);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
static void
bdev_blob_readv(struct spdk_bs_dev *dev, struct spdk_io_channel *channel,
struct iovec *iov, int iovcnt,
uint64_t lba, uint32_t lba_count, struct spdk_bs_dev_cb_args *cb_args)
{
int rc;
rc = spdk_bdev_readv_blocks(__get_desc(dev), channel, iov, iovcnt, lba,
lba_count, bdev_blob_io_complete, cb_args);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, iov, iovcnt, lba, 0,
lba_count, SPDK_BDEV_IO_TYPE_READ, cb_args, NULL);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
static void
bdev_blob_writev(struct spdk_bs_dev *dev, struct spdk_io_channel *channel,
struct iovec *iov, int iovcnt,
uint64_t lba, uint32_t lba_count, struct spdk_bs_dev_cb_args *cb_args)
{
int rc;
rc = spdk_bdev_writev_blocks(__get_desc(dev), channel, iov, iovcnt, lba,
lba_count, bdev_blob_io_complete, cb_args);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, iov, iovcnt, lba, 0,
lba_count, SPDK_BDEV_IO_TYPE_WRITE, cb_args, NULL);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
bdev: remove spdk_bdev_ext_io_opts from spdk_bdev_io The spdk_bdev_ext_io_opts structure is used to pass extra options when submitting a bdev IO request, without having to modify/add functions to handle new options. Additionally, the structure has a size field to allow adding new fields without breaking the ABI (and thus having to bump up the major version of a library). It is also a part of spdk_bdev_io and there are several reasons for removing it from that structure: 1. The size field only makes sense in structures that are passed through pointers. And spdk_bdev_ext_io_opts is indeed passed as a pointer to spdk_bdev_{readv,writev}_blocks_ext(), however it is also embedded in spdk_bdev_io (internal.ext_opts_copy), which is also part of the API. It means that each time a new field is added to spdk_bdev_ext_io_opts, the size of spdk_bdev_io will also change, so we will need to bump the major version of libspdk_bdev anyway, thus making spdk_bdev_ext_io_opts.size useless. 2. The size field also makes internal.ext_opts cumbersome to use, as each time one of its fields is accessed, we need to check the size. Currently the code doesn't do that, because all of the existing spdk_bdev_ext_io_opts fields were present when this structure was initially introduced, but we'd need to do check the size before accessing any new fields. 3. spdk_bdev_ext_io_opts has a metadata field, while spdk_bdev_io already has u.bdev.md_buf, which means that we store the same thing in several different places in spdk_bdev_io (u.bdev.md_buf, u.bdev.ext_opts->metadata, internal.ext_opts->metadata). Therefore, this patch removes all references to spdk_bdev_ext_io_opts from spdk_bdev_io and replaces them with fields (memory_domain, memory_domain_ctx) that were missing in spdk_bdev_io. Unfortunately, this change breaks the API and requires changes in bdev modules that supported spdk_bdev_io.u.bdev.ext_opts. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: I49b7524eb84d1d4d7f12b7ab025fec36da1ee01f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16773 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2023-02-10 15:22:51 +00:00
static inline void
blob_ext_io_opts_to_bdev_opts(struct spdk_bdev_ext_io_opts *dst, struct spdk_blob_ext_io_opts *src)
{
memset(dst, 0, sizeof(*dst));
dst->size = sizeof(*dst);
dst->memory_domain = src->memory_domain;
dst->memory_domain_ctx = src->memory_domain_ctx;
}
static void
bdev_blob_readv_ext(struct spdk_bs_dev *dev, struct spdk_io_channel *channel,
struct iovec *iov, int iovcnt,
uint64_t lba, uint32_t lba_count, struct spdk_bs_dev_cb_args *cb_args,
struct spdk_blob_ext_io_opts *io_opts)
{
bdev: remove spdk_bdev_ext_io_opts from spdk_bdev_io The spdk_bdev_ext_io_opts structure is used to pass extra options when submitting a bdev IO request, without having to modify/add functions to handle new options. Additionally, the structure has a size field to allow adding new fields without breaking the ABI (and thus having to bump up the major version of a library). It is also a part of spdk_bdev_io and there are several reasons for removing it from that structure: 1. The size field only makes sense in structures that are passed through pointers. And spdk_bdev_ext_io_opts is indeed passed as a pointer to spdk_bdev_{readv,writev}_blocks_ext(), however it is also embedded in spdk_bdev_io (internal.ext_opts_copy), which is also part of the API. It means that each time a new field is added to spdk_bdev_ext_io_opts, the size of spdk_bdev_io will also change, so we will need to bump the major version of libspdk_bdev anyway, thus making spdk_bdev_ext_io_opts.size useless. 2. The size field also makes internal.ext_opts cumbersome to use, as each time one of its fields is accessed, we need to check the size. Currently the code doesn't do that, because all of the existing spdk_bdev_ext_io_opts fields were present when this structure was initially introduced, but we'd need to do check the size before accessing any new fields. 3. spdk_bdev_ext_io_opts has a metadata field, while spdk_bdev_io already has u.bdev.md_buf, which means that we store the same thing in several different places in spdk_bdev_io (u.bdev.md_buf, u.bdev.ext_opts->metadata, internal.ext_opts->metadata). Therefore, this patch removes all references to spdk_bdev_ext_io_opts from spdk_bdev_io and replaces them with fields (memory_domain, memory_domain_ctx) that were missing in spdk_bdev_io. Unfortunately, this change breaks the API and requires changes in bdev modules that supported spdk_bdev_io.u.bdev.ext_opts. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: I49b7524eb84d1d4d7f12b7ab025fec36da1ee01f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16773 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2023-02-10 15:22:51 +00:00
struct spdk_bdev_ext_io_opts bdev_io_opts;
int rc;
bdev: remove spdk_bdev_ext_io_opts from spdk_bdev_io The spdk_bdev_ext_io_opts structure is used to pass extra options when submitting a bdev IO request, without having to modify/add functions to handle new options. Additionally, the structure has a size field to allow adding new fields without breaking the ABI (and thus having to bump up the major version of a library). It is also a part of spdk_bdev_io and there are several reasons for removing it from that structure: 1. The size field only makes sense in structures that are passed through pointers. And spdk_bdev_ext_io_opts is indeed passed as a pointer to spdk_bdev_{readv,writev}_blocks_ext(), however it is also embedded in spdk_bdev_io (internal.ext_opts_copy), which is also part of the API. It means that each time a new field is added to spdk_bdev_ext_io_opts, the size of spdk_bdev_io will also change, so we will need to bump the major version of libspdk_bdev anyway, thus making spdk_bdev_ext_io_opts.size useless. 2. The size field also makes internal.ext_opts cumbersome to use, as each time one of its fields is accessed, we need to check the size. Currently the code doesn't do that, because all of the existing spdk_bdev_ext_io_opts fields were present when this structure was initially introduced, but we'd need to do check the size before accessing any new fields. 3. spdk_bdev_ext_io_opts has a metadata field, while spdk_bdev_io already has u.bdev.md_buf, which means that we store the same thing in several different places in spdk_bdev_io (u.bdev.md_buf, u.bdev.ext_opts->metadata, internal.ext_opts->metadata). Therefore, this patch removes all references to spdk_bdev_ext_io_opts from spdk_bdev_io and replaces them with fields (memory_domain, memory_domain_ctx) that were missing in spdk_bdev_io. Unfortunately, this change breaks the API and requires changes in bdev modules that supported spdk_bdev_io.u.bdev.ext_opts. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: I49b7524eb84d1d4d7f12b7ab025fec36da1ee01f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16773 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2023-02-10 15:22:51 +00:00
blob_ext_io_opts_to_bdev_opts(&bdev_io_opts, io_opts);
rc = spdk_bdev_readv_blocks_ext(__get_desc(dev), channel, iov, iovcnt, lba, lba_count,
bdev: remove spdk_bdev_ext_io_opts from spdk_bdev_io The spdk_bdev_ext_io_opts structure is used to pass extra options when submitting a bdev IO request, without having to modify/add functions to handle new options. Additionally, the structure has a size field to allow adding new fields without breaking the ABI (and thus having to bump up the major version of a library). It is also a part of spdk_bdev_io and there are several reasons for removing it from that structure: 1. The size field only makes sense in structures that are passed through pointers. And spdk_bdev_ext_io_opts is indeed passed as a pointer to spdk_bdev_{readv,writev}_blocks_ext(), however it is also embedded in spdk_bdev_io (internal.ext_opts_copy), which is also part of the API. It means that each time a new field is added to spdk_bdev_ext_io_opts, the size of spdk_bdev_io will also change, so we will need to bump the major version of libspdk_bdev anyway, thus making spdk_bdev_ext_io_opts.size useless. 2. The size field also makes internal.ext_opts cumbersome to use, as each time one of its fields is accessed, we need to check the size. Currently the code doesn't do that, because all of the existing spdk_bdev_ext_io_opts fields were present when this structure was initially introduced, but we'd need to do check the size before accessing any new fields. 3. spdk_bdev_ext_io_opts has a metadata field, while spdk_bdev_io already has u.bdev.md_buf, which means that we store the same thing in several different places in spdk_bdev_io (u.bdev.md_buf, u.bdev.ext_opts->metadata, internal.ext_opts->metadata). Therefore, this patch removes all references to spdk_bdev_ext_io_opts from spdk_bdev_io and replaces them with fields (memory_domain, memory_domain_ctx) that were missing in spdk_bdev_io. Unfortunately, this change breaks the API and requires changes in bdev modules that supported spdk_bdev_io.u.bdev.ext_opts. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: I49b7524eb84d1d4d7f12b7ab025fec36da1ee01f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16773 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2023-02-10 15:22:51 +00:00
bdev_blob_io_complete, cb_args, &bdev_io_opts);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, iov, iovcnt, lba, 0, lba_count, SPDK_BDEV_IO_TYPE_READ, cb_args,
io_opts);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
static void
bdev_blob_writev_ext(struct spdk_bs_dev *dev, struct spdk_io_channel *channel,
struct iovec *iov, int iovcnt,
uint64_t lba, uint32_t lba_count, struct spdk_bs_dev_cb_args *cb_args,
struct spdk_blob_ext_io_opts *io_opts)
{
bdev: remove spdk_bdev_ext_io_opts from spdk_bdev_io The spdk_bdev_ext_io_opts structure is used to pass extra options when submitting a bdev IO request, without having to modify/add functions to handle new options. Additionally, the structure has a size field to allow adding new fields without breaking the ABI (and thus having to bump up the major version of a library). It is also a part of spdk_bdev_io and there are several reasons for removing it from that structure: 1. The size field only makes sense in structures that are passed through pointers. And spdk_bdev_ext_io_opts is indeed passed as a pointer to spdk_bdev_{readv,writev}_blocks_ext(), however it is also embedded in spdk_bdev_io (internal.ext_opts_copy), which is also part of the API. It means that each time a new field is added to spdk_bdev_ext_io_opts, the size of spdk_bdev_io will also change, so we will need to bump the major version of libspdk_bdev anyway, thus making spdk_bdev_ext_io_opts.size useless. 2. The size field also makes internal.ext_opts cumbersome to use, as each time one of its fields is accessed, we need to check the size. Currently the code doesn't do that, because all of the existing spdk_bdev_ext_io_opts fields were present when this structure was initially introduced, but we'd need to do check the size before accessing any new fields. 3. spdk_bdev_ext_io_opts has a metadata field, while spdk_bdev_io already has u.bdev.md_buf, which means that we store the same thing in several different places in spdk_bdev_io (u.bdev.md_buf, u.bdev.ext_opts->metadata, internal.ext_opts->metadata). Therefore, this patch removes all references to spdk_bdev_ext_io_opts from spdk_bdev_io and replaces them with fields (memory_domain, memory_domain_ctx) that were missing in spdk_bdev_io. Unfortunately, this change breaks the API and requires changes in bdev modules that supported spdk_bdev_io.u.bdev.ext_opts. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: I49b7524eb84d1d4d7f12b7ab025fec36da1ee01f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16773 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2023-02-10 15:22:51 +00:00
struct spdk_bdev_ext_io_opts bdev_io_opts;
int rc;
bdev: remove spdk_bdev_ext_io_opts from spdk_bdev_io The spdk_bdev_ext_io_opts structure is used to pass extra options when submitting a bdev IO request, without having to modify/add functions to handle new options. Additionally, the structure has a size field to allow adding new fields without breaking the ABI (and thus having to bump up the major version of a library). It is also a part of spdk_bdev_io and there are several reasons for removing it from that structure: 1. The size field only makes sense in structures that are passed through pointers. And spdk_bdev_ext_io_opts is indeed passed as a pointer to spdk_bdev_{readv,writev}_blocks_ext(), however it is also embedded in spdk_bdev_io (internal.ext_opts_copy), which is also part of the API. It means that each time a new field is added to spdk_bdev_ext_io_opts, the size of spdk_bdev_io will also change, so we will need to bump the major version of libspdk_bdev anyway, thus making spdk_bdev_ext_io_opts.size useless. 2. The size field also makes internal.ext_opts cumbersome to use, as each time one of its fields is accessed, we need to check the size. Currently the code doesn't do that, because all of the existing spdk_bdev_ext_io_opts fields were present when this structure was initially introduced, but we'd need to do check the size before accessing any new fields. 3. spdk_bdev_ext_io_opts has a metadata field, while spdk_bdev_io already has u.bdev.md_buf, which means that we store the same thing in several different places in spdk_bdev_io (u.bdev.md_buf, u.bdev.ext_opts->metadata, internal.ext_opts->metadata). Therefore, this patch removes all references to spdk_bdev_ext_io_opts from spdk_bdev_io and replaces them with fields (memory_domain, memory_domain_ctx) that were missing in spdk_bdev_io. Unfortunately, this change breaks the API and requires changes in bdev modules that supported spdk_bdev_io.u.bdev.ext_opts. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: I49b7524eb84d1d4d7f12b7ab025fec36da1ee01f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16773 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2023-02-10 15:22:51 +00:00
blob_ext_io_opts_to_bdev_opts(&bdev_io_opts, io_opts);
rc = spdk_bdev_writev_blocks_ext(__get_desc(dev), channel, iov, iovcnt, lba, lba_count,
bdev: remove spdk_bdev_ext_io_opts from spdk_bdev_io The spdk_bdev_ext_io_opts structure is used to pass extra options when submitting a bdev IO request, without having to modify/add functions to handle new options. Additionally, the structure has a size field to allow adding new fields without breaking the ABI (and thus having to bump up the major version of a library). It is also a part of spdk_bdev_io and there are several reasons for removing it from that structure: 1. The size field only makes sense in structures that are passed through pointers. And spdk_bdev_ext_io_opts is indeed passed as a pointer to spdk_bdev_{readv,writev}_blocks_ext(), however it is also embedded in spdk_bdev_io (internal.ext_opts_copy), which is also part of the API. It means that each time a new field is added to spdk_bdev_ext_io_opts, the size of spdk_bdev_io will also change, so we will need to bump the major version of libspdk_bdev anyway, thus making spdk_bdev_ext_io_opts.size useless. 2. The size field also makes internal.ext_opts cumbersome to use, as each time one of its fields is accessed, we need to check the size. Currently the code doesn't do that, because all of the existing spdk_bdev_ext_io_opts fields were present when this structure was initially introduced, but we'd need to do check the size before accessing any new fields. 3. spdk_bdev_ext_io_opts has a metadata field, while spdk_bdev_io already has u.bdev.md_buf, which means that we store the same thing in several different places in spdk_bdev_io (u.bdev.md_buf, u.bdev.ext_opts->metadata, internal.ext_opts->metadata). Therefore, this patch removes all references to spdk_bdev_ext_io_opts from spdk_bdev_io and replaces them with fields (memory_domain, memory_domain_ctx) that were missing in spdk_bdev_io. Unfortunately, this change breaks the API and requires changes in bdev modules that supported spdk_bdev_io.u.bdev.ext_opts. Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com> Change-Id: I49b7524eb84d1d4d7f12b7ab025fec36da1ee01f Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16773 Tested-by: SPDK CI Jenkins <sys_sgci@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com> Reviewed-by: Jim Harris <james.r.harris@intel.com>
2023-02-10 15:22:51 +00:00
bdev_blob_io_complete, cb_args, &bdev_io_opts);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, iov, iovcnt, lba, 0, lba_count, SPDK_BDEV_IO_TYPE_WRITE, cb_args,
io_opts);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
static void
bdev_blob_write_zeroes(struct spdk_bs_dev *dev, struct spdk_io_channel *channel, uint64_t lba,
blob: use uint64_t for unmap and write_zeroes lba count Previous patches (5363eb3c) tried to work around the 32-bit unmap and write_zeroes LBA counts by breaking up larger operations into smaller chunks of max size UINT32_MAX lba chunks. But some SSDs may just ignore unmap operations that are not aligned to full physical block boundaries - and a UINT32_MAX lba unmap on a 512B logical / 4KiB physical SSD would not be aligned. If the SSD decided to ignore the unmap/deallocate (which it is allowed to do according to NVMe spec), we could end up with not unmapping *any* blocks. Probably SSDs should always be trying hard to unmap as many blocks as possible, but let's not try to depend on that in blobstore. So one option would be to break them into chunks close to UINT32_MAX which are still aligned to 4KiB boundaries. But the better fix is to just change the unmap and write_zeroes APIs to take 64-bit arguments, and then we can avoid the chunking altogether. Fixes issue #2190. Signed-off-by: Jim Harris <james.r.harris@intel.com> Change-Id: I23998e493a764d466927c3520c7a8c7f943000a6 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9737 Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com> Community-CI: Mellanox Build Bot Reviewed-by: Xiaodong Liu <xiaodong.liu@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com> Reviewed-by: Dong Yi <dongx.yi@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
2021-10-05 23:37:19 +00:00
uint64_t lba_count, struct spdk_bs_dev_cb_args *cb_args)
{
int rc;
rc = spdk_bdev_write_zeroes_blocks(__get_desc(dev), channel, lba,
lba_count, bdev_blob_io_complete, cb_args);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, NULL, 0, lba, 0,
lba_count, SPDK_BDEV_IO_TYPE_WRITE_ZEROES, cb_args, NULL);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
static void
bdev_blob_unmap(struct spdk_bs_dev *dev, struct spdk_io_channel *channel, uint64_t lba,
blob: use uint64_t for unmap and write_zeroes lba count Previous patches (5363eb3c) tried to work around the 32-bit unmap and write_zeroes LBA counts by breaking up larger operations into smaller chunks of max size UINT32_MAX lba chunks. But some SSDs may just ignore unmap operations that are not aligned to full physical block boundaries - and a UINT32_MAX lba unmap on a 512B logical / 4KiB physical SSD would not be aligned. If the SSD decided to ignore the unmap/deallocate (which it is allowed to do according to NVMe spec), we could end up with not unmapping *any* blocks. Probably SSDs should always be trying hard to unmap as many blocks as possible, but let's not try to depend on that in blobstore. So one option would be to break them into chunks close to UINT32_MAX which are still aligned to 4KiB boundaries. But the better fix is to just change the unmap and write_zeroes APIs to take 64-bit arguments, and then we can avoid the chunking altogether. Fixes issue #2190. Signed-off-by: Jim Harris <james.r.harris@intel.com> Change-Id: I23998e493a764d466927c3520c7a8c7f943000a6 Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9737 Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com> Community-CI: Mellanox Build Bot Reviewed-by: Xiaodong Liu <xiaodong.liu@intel.com> Reviewed-by: Changpeng Liu <changpeng.liu@intel.com> Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com> Reviewed-by: Dong Yi <dongx.yi@intel.com> Reviewed-by: Ben Walker <benjamin.walker@intel.com> Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com> Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
2021-10-05 23:37:19 +00:00
uint64_t lba_count, struct spdk_bs_dev_cb_args *cb_args)
{
struct blob_bdev *blob_bdev = (struct blob_bdev *)dev;
int rc;
if (spdk_bdev_io_type_supported(blob_bdev->bdev, SPDK_BDEV_IO_TYPE_UNMAP)) {
rc = spdk_bdev_unmap_blocks(__get_desc(dev), channel, lba, lba_count,
bdev_blob_io_complete, cb_args);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, NULL, 0, lba, 0,
lba_count, SPDK_BDEV_IO_TYPE_UNMAP, cb_args, NULL);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
} else {
/*
* If the device doesn't support unmap, immediately complete
* the request. Blobstore does not rely on unmap zeroing
* data.
*/
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, 0);
}
}
static void
bdev_blob_copy(struct spdk_bs_dev *dev, struct spdk_io_channel *channel,
uint64_t dst_lba, uint64_t src_lba, uint64_t lba_count,
struct spdk_bs_dev_cb_args *cb_args)
{
int rc;
rc = spdk_bdev_copy_blocks(__get_desc(dev), channel,
dst_lba, src_lba, lba_count,
bdev_blob_io_complete, cb_args);
if (rc == -ENOMEM) {
bdev_blob_queue_io(dev, channel, NULL, 0, dst_lba, src_lba,
lba_count, SPDK_BDEV_IO_TYPE_COPY, cb_args, NULL);
} else if (rc != 0) {
cb_args->cb_fn(cb_args->channel, cb_args->cb_arg, rc);
}
}
static void
bdev_blob_resubmit(void *arg)
{
struct blob_resubmit *ctx = (struct blob_resubmit *) arg;
switch (ctx->io_type) {
case SPDK_BDEV_IO_TYPE_READ:
if (ctx->iovcnt > 0) {
bdev_blob_readv_ext(ctx->dev, ctx->channel, (struct iovec *) ctx->payload, ctx->iovcnt,
ctx->lba, ctx->lba_count, ctx->cb_args, ctx->ext_io_opts);
} else {
bdev_blob_read(ctx->dev, ctx->channel, ctx->payload,
ctx->lba, ctx->lba_count, ctx->cb_args);
}
break;
case SPDK_BDEV_IO_TYPE_WRITE:
if (ctx->iovcnt > 0) {
bdev_blob_writev_ext(ctx->dev, ctx->channel, (struct iovec *) ctx->payload, ctx->iovcnt,
ctx->lba, ctx->lba_count, ctx->cb_args, ctx->ext_io_opts);
} else {
bdev_blob_write(ctx->dev, ctx->channel, ctx->payload,
ctx->lba, ctx->lba_count, ctx->cb_args);
}
break;
case SPDK_BDEV_IO_TYPE_UNMAP:
bdev_blob_unmap(ctx->dev, ctx->channel,
ctx->lba, ctx->lba_count, ctx->cb_args);
break;
case SPDK_BDEV_IO_TYPE_WRITE_ZEROES:
bdev_blob_write_zeroes(ctx->dev, ctx->channel,
ctx->lba, ctx->lba_count, ctx->cb_args);
break;
case SPDK_BDEV_IO_TYPE_COPY:
bdev_blob_copy(ctx->dev, ctx->channel,
ctx->lba, ctx->src_lba, ctx->lba_count, ctx->cb_args);
break;
default:
SPDK_ERRLOG("Unsupported io type %d\n", ctx->io_type);
assert(false);
break;
}
free(ctx);
}
int
spdk_bs_bdev_claim(struct spdk_bs_dev *bs_dev, struct spdk_bdev_module *module)
{
struct blob_bdev *blob_bdev = (struct blob_bdev *)bs_dev;
struct spdk_bdev_desc *desc = blob_bdev->desc;
enum spdk_bdev_claim_type claim_type;
int rc;
claim_type = blob_bdev->write ? SPDK_BDEV_CLAIM_READ_MANY_WRITE_ONE :
SPDK_BDEV_CLAIM_READ_MANY_WRITE_NONE;
rc = spdk_bdev_module_claim_bdev_desc(desc, claim_type, NULL, module);
if (rc != 0) {
SPDK_ERRLOG("could not claim bs dev\n");
return rc;
}
return rc;
}
static struct spdk_io_channel *
bdev_blob_create_channel(struct spdk_bs_dev *dev)
{
struct blob_bdev *blob_bdev = (struct blob_bdev *)dev;
struct spdk_io_channel *ch;
ch = spdk_bdev_get_io_channel(blob_bdev->desc);
if (ch != NULL) {
spdk_spin_lock(&blob_bdev->lock);
blob_bdev->refs++;
spdk_spin_unlock(&blob_bdev->lock);
}
return ch;
}
static void
bdev_blob_free(struct blob_bdev *blob_bdev)
{
assert(blob_bdev->refs == 0);
spdk_spin_destroy(&blob_bdev->lock);
free(blob_bdev);
}
static void
bdev_blob_destroy_channel(struct spdk_bs_dev *dev, struct spdk_io_channel *channel)
{
struct blob_bdev *blob_bdev = (struct blob_bdev *)dev;
int32_t refs;
spdk_spin_lock(&blob_bdev->lock);
assert(blob_bdev->refs > 0);
blob_bdev->refs--;
refs = blob_bdev->refs;
spdk_spin_unlock(&blob_bdev->lock);
spdk_put_io_channel(channel);
/*
* If the value of blob_bdev->refs taken while holding blob_bdev->refs is zero, the blob and
* this channel have been destroyed. This means that dev->destroy() has been called and it
* would be an error (akin to use after free) if dev is dereferenced after destroying it.
* Thus, there should be no race with bdev_blob_create_channel().
*
* Because the value of blob_bdev->refs was taken while holding the lock here and the same
* is done in bdev_blob_destroy(), there is no race with bdev_blob_destroy().
*/
if (refs == 0) {
bdev_blob_free(blob_bdev);
}
}
static void
bdev_blob_destroy(struct spdk_bs_dev *bs_dev)
{
struct blob_bdev *blob_bdev = (struct blob_bdev *)bs_dev;
struct spdk_bdev_desc *desc;
int32_t refs;
spdk_spin_lock(&blob_bdev->lock);
desc = blob_bdev->desc;
blob_bdev->desc = NULL;
blob_bdev->refs--;
refs = blob_bdev->refs;
spdk_spin_unlock(&blob_bdev->lock);
spdk_bdev_close(desc);
/*
* If the value of blob_bdev->refs taken while holding blob_bdev->refs is zero,
* bs_dev->destroy() has been called and all the channels have been destroyed. It would be
* an error (akin to use after free) if bs_dev is dereferenced after destroying it. Thus,
* there should be no race with bdev_blob_create_channel().
*
* Because the value of blob_bdev->refs was taken while holding the lock here and the same
* is done in bdev_blob_destroy_channel(), there is no race with
* bdev_blob_destroy_channel().
*/
if (refs == 0) {
bdev_blob_free(blob_bdev);
}
}
static struct spdk_bdev *
bdev_blob_get_base_bdev(struct spdk_bs_dev *bs_dev)
{
return __get_bdev(bs_dev);
}
static bool
bdev_blob_is_zeroes(struct spdk_bs_dev *dev, uint64_t lba, uint64_t lba_count)
{
return false;
}
static bool
bdev_blob_translate_lba(struct spdk_bs_dev *dev, uint64_t lba, uint64_t *base_lba)
{
*base_lba = lba;
return true;
}
static void
blob_bdev_init(struct blob_bdev *b, struct spdk_bdev_desc *desc)
{
struct spdk_bdev *bdev;
bdev = spdk_bdev_desc_get_bdev(desc);
assert(bdev != NULL);
b->bdev = bdev;
b->desc = desc;
b->bs_dev.blockcnt = spdk_bdev_get_num_blocks(bdev);
b->bs_dev.blocklen = spdk_bdev_get_block_size(bdev);
b->bs_dev.create_channel = bdev_blob_create_channel;
b->bs_dev.destroy_channel = bdev_blob_destroy_channel;
b->bs_dev.destroy = bdev_blob_destroy;
b->bs_dev.read = bdev_blob_read;
b->bs_dev.write = bdev_blob_write;
b->bs_dev.readv = bdev_blob_readv;
b->bs_dev.writev = bdev_blob_writev;
b->bs_dev.readv_ext = bdev_blob_readv_ext;
b->bs_dev.writev_ext = bdev_blob_writev_ext;
b->bs_dev.write_zeroes = bdev_blob_write_zeroes;
b->bs_dev.unmap = bdev_blob_unmap;
if (spdk_bdev_io_type_supported(bdev, SPDK_BDEV_IO_TYPE_COPY)) {
b->bs_dev.copy = bdev_blob_copy;
}
b->bs_dev.get_base_bdev = bdev_blob_get_base_bdev;
b->bs_dev.is_zeroes = bdev_blob_is_zeroes;
b->bs_dev.translate_lba = bdev_blob_translate_lba;
}
int
spdk_bdev_create_bs_dev(const char *bdev_name, bool write,
struct spdk_bdev_bs_dev_opts *opts, size_t opts_size,
spdk_bdev_event_cb_t event_cb, void *event_ctx,
struct spdk_bs_dev **bs_dev)
{
struct blob_bdev *b;
struct spdk_bdev_desc *desc;
int rc;
assert(spdk_get_thread() != NULL);
if (opts != NULL && opts_size != sizeof(*opts)) {
SPDK_ERRLOG("bdev name '%s': unsupported options\n", bdev_name);
return -EINVAL;
}
b = calloc(1, sizeof(*b));
if (b == NULL) {
SPDK_ERRLOG("could not allocate blob_bdev\n");
return -ENOMEM;
}
rc = spdk_bdev_open_ext(bdev_name, write, event_cb, event_ctx, &desc);
if (rc != 0) {
free(b);
return rc;
}
blob_bdev_init(b, desc);
*bs_dev = &b->bs_dev;
b->write = write;
b->refs = 1;
spdk_spin_init(&b->lock);
return 0;
}
int
spdk_bdev_create_bs_dev_ext(const char *bdev_name, spdk_bdev_event_cb_t event_cb,
void *event_ctx, struct spdk_bs_dev **bs_dev)
{
return spdk_bdev_create_bs_dev(bdev_name, true, NULL, 0, event_cb, event_ctx, bs_dev);
}