Skip to contentSkip to conversation

block/file-posix: Optimize for macOS

PATCH Superseded Akihiko Odaki 2 participants 4 messages Mar 10, 2021 – Apr 1, 2021 Block layer

v3 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v3 1/2 block/file-posix: Optimize for macOS

This commit introduces "punch hole" operation and optimizes transfer block size for macOS.

This commit introduces two additional members, discard_granularity and opt_io to BlockSizes type in include/block/block.h. Also, the members of the type are now optional. Set -1 to discard_granularity and 0 to other members for the default values.

Thanks to Konstantin Nazarov for detailed analysis of a flaw in an old version of this change:
https://gist.github.com/akihikodaki/87df4149e7ca87f18dc56807ec5a1bc5#gistcomment-3654667

---
 block/file-posix.c    | 40 ++++++++++++++++++++++++++++++++++++++--
 block/nvme.c          |  2 ++
 block/raw-format.c    |  4 +++-
 hw/block/block.c      | 12 ++++++++++--
 include/block/block.h |  2 ++
 5 files changed, 55 insertions(+), 5 deletions(-)
block/file-posix.c +38 -2
index 05079b40cae..21bdaf969c5 100644
@@ -44,6 +44,7 @@
 #if defined(__APPLE__) && (__MACH__)
 #include <paths.h>
 #include <sys/param.h>
+#include <sys/mount.h>
 #include <IOKit/IOKitLib.h>
 #include <IOKit/IOBSD.h>
 #include <IOKit/storage/IOMediaBSDClient.h>
@@ -1292,6 +1293,8 @@ static int hdev_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
     if (check_for_dasd(s->fd) < 0) {
         return -ENOTSUP;
     }
+    bsz->opt_io = 0;
+    bsz->discard_granularity = -1;
     ret = probe_logical_blocksize(s->fd, &bsz->log);
     if (ret < 0) {
         return ret;
@@ -1586,6 +1589,7 @@ out:
     }
 }
 
+G_GNUC_UNUSED
 static int translate_err(int err)
 {
     if (err == -ENODEV || err == -ENOSYS || err == -EOPNOTSUPP ||
@@ -1795,16 +1799,27 @@ static int handle_aiocb_discard(void *opaque)
             }
         } while (errno == EINTR);
 
-        ret = -errno;
+        ret = translate_err(-errno);
 #endif
     } else {
 #ifdef CONFIG_FALLOCATE_PUNCH_HOLE
         ret = do_fallocate(s->fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
                            aiocb->aio_offset, aiocb->aio_nbytes);
+        ret = translate_err(-errno);
+#elif defined(__APPLE__) && (__MACH__)
+        fpunchhole_t fpunchhole;
+        fpunchhole.fp_flags = 0;
+        fpunchhole.reserved = 0;
+        fpunchhole.fp_offset = aiocb->aio_offset;
+        fpunchhole.fp_length = aiocb->aio_nbytes;
+        if (fcntl(s->fd, F_PUNCHHOLE, &fpunchhole) == -1) {
+            ret = errno == ENODEV ? -ENOTSUP : -errno;
+        } else {
+            ret = 0;
+        }
 #endif
     }
 
-    ret = translate_err(ret);
     if (ret == -ENOTSUP) {
         s->has_discard = false;
     }
@@ -2113,6 +2128,26 @@ static int raw_co_flush_to_disk(BlockDriverState *bs)
     return raw_thread_pool_submit(bs, handle_aiocb_flush, &acb);
 }
 
+static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
+{
+#if defined(__APPLE__) && (__MACH__)
+    BDRVRawState *s = bs->opaque;
+    struct statfs buf;
+
+    if (!fstatfs(s->fd, &buf)) {
+        bsz->phys = 0;
+        bsz->log = 0;
+        bsz->opt_io = buf.f_iosize;
+        bsz->discard_granularity = buf.f_bsize;
+        return 0;
+    }
+
+    return -errno;
+#else
+    return -ENOTSUP;
+#endif
+}
+
 static void raw_aio_attach_aio_context(BlockDriverState *bs,
                                        AioContext *new_context)
 {
@@ -3247,6 +3282,7 @@ BlockDriver bdrv_file = {
     .bdrv_refresh_limits = raw_refresh_limits,
     .bdrv_io_plug = raw_aio_plug,
     .bdrv_io_unplug = raw_aio_unplug,
+    .bdrv_probe_blocksizes = raw_probe_blocksizes,
     .bdrv_attach_aio_context = raw_aio_attach_aio_context,
 
     .bdrv_co_truncate = raw_co_truncate,
block/nvme.c +2 -0
index 2b5421e7aa6..1845d07577b 100644
@@ -989,6 +989,8 @@ static int nvme_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
     uint32_t blocksize = nvme_get_blocksize(bs);
     bsz->phys = blocksize;
     bsz->log = blocksize;
+    bsz->opt_io = 0;
+    bsz->discard_granularity = -1;
     return 0;
 }
 
block/raw-format.c +3 -1
index 7717578ed6a..847df11f2ae 100644
@@ -507,6 +507,7 @@ static int raw_probe(const uint8_t *buf, int buf_size, const char *filename)
 static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
 {
     BDRVRawState *s = bs->opaque;
+    uint32_t size;
     int ret;
 
     ret = bdrv_probe_blocksizes(bs->file->bs, bsz);
@@ -514,7 +515,8 @@ static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
         return ret;
     }
 
-    if (!QEMU_IS_ALIGNED(s->offset, MAX(bsz->log, bsz->phys))) {
+    size = MAX(bsz->log, bsz->phys);
+    if (size && !QEMU_IS_ALIGNED(s->offset, size)) {
         return -ENOTSUP;
     }
 
hw/block/block.c +10 -2
index 1e34573da71..c907e5a7722 100644
@@ -70,19 +70,27 @@ bool blkconf_blocksizes(BlockConf *conf, Error **errp)
     backend_ret = blk_probe_blocksizes(blk, &blocksizes);
     /* fill in detected values if they are not defined via qemu command line */
     if (!conf->physical_block_size) {
-        if (!backend_ret) {
+        if (!backend_ret && blocksizes.phys) {
            conf->physical_block_size = blocksizes.phys;
         } else {
             conf->physical_block_size = BDRV_SECTOR_SIZE;
         }
     }
     if (!conf->logical_block_size) {
-        if (!backend_ret) {
+        if (!backend_ret && blocksizes.log) {
             conf->logical_block_size = blocksizes.log;
         } else {
             conf->logical_block_size = BDRV_SECTOR_SIZE;
         }
     }
+    if (!backend_ret) {
+        if (!conf->opt_io_size) {
+            conf->opt_io_size = blocksizes.opt_io;
+        }
+        if (conf->discard_granularity == -1) {
+            conf->discard_granularity = blocksizes.discard_granularity;
+        }
+    }
 
     if (conf->logical_block_size > conf->physical_block_size) {
         error_setg(errp,
include/block/block.h +2 -0
index b3f6e509d49..d12471a6cc4 100644
@@ -93,6 +93,8 @@ typedef enum {
 typedef struct BlockSizes {
     uint32_t phys;
     uint32_t log;
+    uint32_t discard_granularity;
+    uint32_t opt_io;
 } BlockSizes;
 
 typedef struct HDGeometry {
Signature · 1 line
2.24.3 (Apple Git-128)
#1 lore raw

PATCH v3 2/2 block: Add backend_defaults property

backend_defaults property allow users to control if default block properties should be decided with backend information.

If it is off, any backend information will be discarded, which is suitable if you plan to perform live migration to a different disk backend.

If it is on, a block device may utilize backend information more aggressively.

By default, it is auto, which uses backend information from physical devices and ignores one from file. It is consistent with the older versions, and should be aligned with the user expectation that a file backend is more independent of host than a physical device backend.

---
 block/file-posix.c       |  2 ++
 block/nvme.c             |  1 +
 hw/block/block.c         | 27 ++++++++++++++++++++++-----
 include/block/block.h    |  1 +
 include/hw/block/block.h |  3 +++
 5 files changed, 29 insertions(+), 5 deletions(-)
block/file-posix.c +2 -0
index 21bdaf969c5..2631bd972a5 100644
@@ -1293,6 +1293,7 @@ static int hdev_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
     if (check_for_dasd(s->fd) < 0) {
         return -ENOTSUP;
     }
+    bsz->file = false;
     bsz->opt_io = 0;
     bsz->discard_granularity = -1;
     ret = probe_logical_blocksize(s->fd, &bsz->log);
@@ -2135,6 +2136,7 @@ static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
     struct statfs buf;
 
     if (!fstatfs(s->fd, &buf)) {
+        bsz->file = true;
         bsz->phys = 0;
         bsz->log = 0;
         bsz->opt_io = buf.f_iosize;
block/nvme.c +1 -0
index 1845d07577b..8438c8a2b90 100644
@@ -987,6 +987,7 @@ static uint32_t nvme_get_blocksize(BlockDriverState *bs)
 static int nvme_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
 {
     uint32_t blocksize = nvme_get_blocksize(bs);
+    bsz->file = false;
     bsz->phys = blocksize;
     bsz->log = blocksize;
     bsz->opt_io = 0;
hw/block/block.c +22 -5
index c907e5a7722..d42cc40a1c8 100644
@@ -65,25 +65,42 @@ bool blkconf_blocksizes(BlockConf *conf, Error **errp)
 {
     BlockBackend *blk = conf->blk;
     BlockSizes blocksizes;
-    int backend_ret;
+    bool use_blocksizes;
+
+    switch (conf->backend_defaults) {
+    case ON_OFF_AUTO_AUTO:
+        use_blocksizes = !blk_probe_blocksizes(blk, &blocksizes) &&
+                         !blocksizes.file;
+        break;
+
+    case ON_OFF_AUTO_ON:
+        use_blocksizes = !blk_probe_blocksizes(blk, &blocksizes);
+        break;
+
+    case ON_OFF_AUTO_OFF:
+        use_blocksizes = false;
+        break;
+
+    default:
+        abort();
+    }
 
-    backend_ret = blk_probe_blocksizes(blk, &blocksizes);
     /* fill in detected values if they are not defined via qemu command line */
     if (!conf->physical_block_size) {
-        if (!backend_ret && blocksizes.phys) {
+        if (use_blocksizes && blocksizes.phys) {
            conf->physical_block_size = blocksizes.phys;
         } else {
             conf->physical_block_size = BDRV_SECTOR_SIZE;
         }
     }
     if (!conf->logical_block_size) {
-        if (!backend_ret && blocksizes.log) {
+        if (use_blocksizes && blocksizes.log) {
             conf->logical_block_size = blocksizes.log;
         } else {
             conf->logical_block_size = BDRV_SECTOR_SIZE;
         }
     }
-    if (!backend_ret) {
+    if (use_blocksizes) {
         if (!conf->opt_io_size) {
             conf->opt_io_size = blocksizes.opt_io;
         }
include/block/block.h +1 -0
index d12471a6cc4..ccc3ec5c847 100644
@@ -91,6 +91,7 @@ typedef enum {
 } BdrvRequestFlags;
 
 typedef struct BlockSizes {
+    bool file;
     uint32_t phys;
     uint32_t log;
     uint32_t discard_granularity;
include/hw/block/block.h +3 -0
index c172cbe65f1..5902c0440a5 100644
@@ -19,6 +19,7 @@
 
 typedef struct BlockConf {
     BlockBackend *blk;
+    OnOffAuto backend_defaults;
     uint32_t physical_block_size;
     uint32_t logical_block_size;
     uint32_t min_io_size;
@@ -48,6 +49,8 @@ static inline unsigned int get_physical_block_exp(BlockConf *conf)
 }
 
 #define DEFINE_BLOCK_PROPERTIES_BASE(_state, _conf)                     \
+    DEFINE_PROP_ON_OFF_AUTO("backend_defaults", _state,                 \
+                            _conf.backend_defaults, ON_OFF_AUTO_AUTO),  \
     DEFINE_PROP_BLOCKSIZE("logical_block_size", _state,                 \
                           _conf.logical_block_size),                    \
     DEFINE_PROP_BLOCKSIZE("physical_block_size", _state,                \
Signature · 1 line
2.24.3 (Apple Git-128)
#2 lore raw

Re: [PATCH v3 2/2] block: Add backend_defaults property

Quoted 6 lines · Akihiko Odaki

backend_defaults property allow users to control if default block properties should be decided with backend information.

If it is off, any backend information will be discarded, which is suitable if you plan to perform live migration to a different disk backend.

Quoted 7 lines

If it is on, a block device may utilize backend information more aggressively.

By default, it is auto, which uses backend information from physical devices and ignores one from file. It is consistent with the older versions, and should be aligned with the user expectation that a file backend is more independent of host than a physical device backend.

Can BlockLimits pdiscard_alignment and opt_transfer be used instead of adding discard_granularity and opt_io fields to BlockSizes? They seem to do the same thing except the QEMU block layer currently uses BlockLimits values internally to ensure that requests are suitable for the host device.

#3 lore raw
In reply to Akihiko Odaki

block/file-posix.c out: · line 1592

Quoted 5 lines · Akihiko Odaki
block/file-posix.c +1 -0
@@ -1586,6 +1589,7 @@ out:
     }
 }
 
+G_GNUC_UNUSED

Why isn't translate_err() used in the F_PUNCHHOLE case below?

If you really want to avoid using it on macOS, please add a #if with the necessary conditions here so it's clear when this translate_err() is needed.

block/raw-format.c static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz) · line 522

Quoted 10 lines
block/raw-format.c +2 -1
@@ -514,7 +515,8 @@ static int raw_probe_blocksizes(BlockDriverState *bs, BlockSizes *bsz)
         return ret;
     }
 
-    if (!QEMU_IS_ALIGNED(s->offset, MAX(bsz->log, bsz->phys))) {
+    size = MAX(bsz->log, bsz->phys);
+    if (size && !QEMU_IS_ALIGNED(s->offset, size)) {
         return -ENOTSUP;
     }
 

This patch changes the semantics of bdrv_probe_blocksizes(). It used to return -ENOTSUP when phys/log weren't available. Now it returns 0 and the fields are 0. Please update the bdrv_probe_blocksizes doc comment in include/block/block_int.h to mention phys and log, as well as that fields can be set to 0 (or -1 in the case of discard_granularity).