On Sun, 2022-10-16 at 22:56 +0800, Sam Li wrote:
> Since Linux doesn't have a user API to issue zone append operations to
> zoned devices from user space, the file-posix driver is modified to add
> zone append emulation using regular writes. To do this, the file-posix
> driver tracks the wp location of all zones of the device. It uses an
> array of uint64_t. The most significant bit of each wp location indicates
> if the zone type is conventional zones.
> 
> The zones wp can be changed due to the following operations issued:
> - zone reset: change the wp to the start offset of that zone
> - zone finish: change to the end location of that zone
> - write to a zone
> - zone append
> 
> Signed-off-by: Sam Li <faithilike...@gmail.com>
> ---
>  block/file-posix.c               | 144 +++++++++++++++++++++++++++++++
>  include/block/block-common.h     |  14 +++
>  include/block/block_int-common.h |   3 +
>  3 files changed, 161 insertions(+)
> 
> diff --git a/block/file-posix.c b/block/file-posix.c
> index 7c5a330fc1..5ff5500301 100644
> --- a/block/file-posix.c
> +++ b/block/file-posix.c
> @@ -1324,6 +1324,66 @@ static int hdev_get_max_segments(int fd, struct stat
> *st)
>  #endif
>  }
>  
> +#if defined(CONFIG_BLKZONED)
> +static int get_zones_wp(int fd, BlockZoneWps *wps, int64_t offset,
> +                        unsigned int nrz) {
> +    struct blk_zone *blkz;
> +    int64_t rep_size;

size_t

> +    int64_t sector = offset >> BDRV_SECTOR_BITS;

uint64_t

> +    int ret, n = 0, i = 0;
> +    rep_size = sizeof(struct blk_zone_report) + nrz * sizeof(struct 
> blk_zone);
> +    g_autofree struct blk_zone_report *rep = NULL;
> +
> +    rep = g_malloc(rep_size);
> +    blkz = (struct blk_zone *)(rep + 1);
> +    while (n < nrz) {
> +        memset(rep, 0, rep_size);
> +        rep->sector = sector;
> +        rep->nr_zones = nrz - n;
> +
> +        do {
> +            ret = ioctl(fd, BLKREPORTZONE, rep);
> +        } while (ret != 0 && errno == EINTR);
> +        if (ret != 0) {
> +            error_report("%d: ioctl BLKREPORTZONE at %" PRId64 " failed %d",
> +                    fd, offset, errno);
> +            return -errno;
> +        }
> +
> +        if (!rep->nr_zones) {
> +            break;
> +        }
> +
> +        for (i = 0; i < rep->nr_zones; i++, n++) {
> +            /*
> +             * The wp tracking cares only about sequential writes required 
> and
> +             * sequential write preferred zones so that the wp can advance to
> +             * the right location.
> +             * Use the most significant bit of the wp location to indicate 
> the
> +             * zone type: 0 for SWR/SWP zones and 1 for conventional zones.
> +             */
> +            if (blkz[i].type == BLK_ZONE_TYPE_CONVENTIONAL) {
> +                wps->wp[i] = 1ULL << 63;
> +            } else {
> +                wps->wp[i] = blkz[i].wp << BDRV_SECTOR_BITS;
> +            }
> +        }
> +        sector = blkz[i - 1].start + blkz[i - 1].len;
> +    }
> +
> +    return 0;
> +}
> +
> +static void update_zones_wp(int fd, BlockZoneWps *wps, int64_t offset,
> +                            unsigned int nrz) {
> +    qemu_mutex_lock(&wps->lock);
> +    if (get_zones_wp(fd, wps, offset, nrz) < 0) {
> +        error_report("update zone wp failed");
> +    }
> +    qemu_mutex_unlock(&wps->lock);
> +}
> +#endif
> +
>  static void raw_refresh_limits(BlockDriverState *bs, Error **errp)
>  {
>      BDRVRawState *s = bs->opaque;
> @@ -1414,6 +1474,14 @@ static void raw_refresh_limits(BlockDriverState *bs,
> Error **errp)
>          if (ret >= 0) {
>              bs->bl.max_active_zones = ret;
>          }
> +
> +        bs->bl.wps = g_malloc(sizeof(BlockZoneWps) + sizeof(int64_t) * ret);
> +        if (get_zones_wp(s->fd, bs->bl.wps, 0, ret) < 0) {
> +            error_report("report wps failed");
> +            g_free(bs->bl.wps);
> +            return;
> +        }
> +        qemu_mutex_init(&bs->bl.wps->lock);
>      }
>  }
>  
> @@ -1725,6 +1793,25 @@ static int handle_aiocb_rw(void *opaque)
>  
>  out:
>      if (nbytes == aiocb->aio_nbytes) {
> +#if defined(CONFIG_BLKZONED)
> +        if (aiocb->aio_type & QEMU_AIO_WRITE) {
> +            BlockZoneWps *wps = aiocb->bs->bl.wps;
> +            int index = aiocb->aio_offset / aiocb->bs->bl.zone_size;
> +            if (wps) {

In my testing, I get a divide by zero exception in the "index"
calculation above. Changing this part as follows

-            int index = aiocb->aio_offset / aiocb->bs->bl.zone_size;
-            if (wps) {
+            if (wps && aiocb->bs->bl.zone_size) {
+                int index = aiocb->aio_offset / aiocb->bs->bl.zone_size;
+

fixes the crash.

> +                qemu_mutex_lock(&wps->lock);
> +                if (!BDRV_ZT_IS_CONV(wps->wp[index])) {
> +                    uint64_t wend_offset =
> +                            aiocb->aio_offset + aiocb->aio_nbytes;
> +
> +                    /* Advance the wp if needed */
> +                    if (wend_offset > wps->wp[index]) {
> +                        wps->wp[index] = wend_offset;
> +                    }
> +                }
> +                qemu_mutex_unlock(&wps->lock);
> +            }
> +        }
> +#endif
>          return 0;
>      } else if (nbytes >= 0 && nbytes < aiocb->aio_nbytes) {
>          if (aiocb->aio_type & QEMU_AIO_WRITE) {
> @@ -1736,6 +1823,11 @@ out:
>          }
>      } else {
>          assert(nbytes < 0);
> +#if defined(CONFIG_BLKZONED)
> +        if (aiocb->aio_type & QEMU_AIO_WRITE) {
> +            update_zones_wp(aiocb->aio_fildes, aiocb->bs->bl.wps, 0, 1);
> +        }
> +#endif
>          return nbytes;
>      }
>  }
> @@ -2022,14 +2114,29 @@ static int handle_aiocb_zone_report(void *opaque)
>  #endif
>  
>  #if defined(CONFIG_BLKZONED)
> +static bool zone_is_empty(BlockDriverState *bs, int64_t index)
> +{
> +    return bs->bl.wps->wp[index] == index * bs->bl.zone_size;
> +}
> +
>  static int handle_aiocb_zone_mgmt(void *opaque)
>  {
>      RawPosixAIOData *aiocb = opaque;
> +    BlockDriverState *bs = aiocb->bs;
>      int fd = aiocb->aio_fildes;
>      int64_t sector = aiocb->aio_offset / 512;
>      int64_t nr_sectors = aiocb->aio_nbytes / 512;
>      struct blk_zone_range range;
>      int ret;
> +    uint64_t wend_offset;
> +    BlockZoneWps *wps = bs->bl.wps;
> +    int index = aiocb->aio_offset / bs->bl.zone_size;

uint32_t

> +
> +    if (BDRV_ZT_IS_CONV(wps->wp[index]) &&
> +        aiocb->aio_nbytes != bs->bl.capacity) {
> +        error_report("zone mgmt operations not allowed for conventional
> zones");
> +        return -EIO;
> +    }
>  
>      /* Execute the operation */
>      range.sector = sector;
> @@ -2038,11 +2145,42 @@ static int handle_aiocb_zone_mgmt(void *opaque)
>          ret = ioctl(fd, aiocb->zone_mgmt.zone_op, &range);
>      } while (ret != 0 && errno == EINTR);
>      if (ret != 0) {
> +        update_zones_wp(aiocb->aio_fildes, aiocb->bs->bl.wps,
> +                        aiocb->aio_offset,
> +                        aiocb->aio_nbytes / bs->bl.zone_size);
>          ret = -errno;
>          error_report("ioctl %s failed %d", aiocb->zone_mgmt.zone_op_name,
>                       ret);
>          return ret;
>      }
> +    qemu_mutex_lock(&wps->lock);
> +    if (!BDRV_ZT_IS_CONV(wps->wp[index])) {
> +        /*
> +         * The zoned device allows the last zone smaller that the zone size.
> +         */
> +        if (aiocb->aio_nbytes < bs->bl.zone_size) {
> +            wend_offset = aiocb->aio_offset + aiocb->aio_nbytes;
> +        } else {
> +            wend_offset = aiocb->aio_offset + bs->bl.zone_size;
> +        }
> +
> +        if (!zone_is_empty(bs, index) &&
> +            aiocb->zone_mgmt.zone_op == BLKRESETZONE) {
> +            if (aiocb->aio_nbytes == bs->bl.capacity) {
> +                for (int i = 0; i < bs->bl.nr_zones; ++i) {
> +                    if (!BDRV_ZT_IS_CONV(wps->wp[index])) {
> +                        wps->wp[i] = i * bs->bl.zone_size;
> +                    }
> +                }
> +            } else {
> +                wps->wp[index] = aiocb->aio_offset;
> +            }
> +        } else if (aiocb->zone_mgmt.zone_op == BLKFINISHZONE) {
> +            wps->wp[index] = wend_offset;
> +        }
> +    }
> +    qemu_mutex_unlock(&wps->lock);
> +
>      return 0;
>  }
>  #endif
> @@ -2483,6 +2621,12 @@ static void raw_close(BlockDriverState *bs)
>      BDRVRawState *s = bs->opaque;
>  
>      if (s->fd >= 0) {
> +#if defined(CONFIG_BLKZONED)
> +        if (bs->bl.wps) {
> +            qemu_mutex_destroy(&bs->bl.wps->lock);
> +            g_free(bs->bl.wps);
> +        }
> +#endif
>          qemu_close(s->fd);
>          s->fd = -1;
>      }
> diff --git a/include/block/block-common.h b/include/block/block-common.h
> index 36bd0e480e..52372f8252 100644
> --- a/include/block/block-common.h
> +++ b/include/block/block-common.h
> @@ -92,6 +92,14 @@ typedef struct BlockZoneDescriptor {
>      BlockZoneCondition cond;
>  } BlockZoneDescriptor;
>  
> +/*
> + * Track write pointers of a zone in bytes.
> + */
> +typedef struct BlockZoneWps {
> +    QemuMutex lock;
> +    uint64_t wp[];
> +} BlockZoneWps;
> +
>  typedef struct BlockDriverInfo {
>      /* in bytes, 0 if irrelevant */
>      int cluster_size;
> @@ -205,6 +213,12 @@ typedef enum {
>  #define BDRV_SECTOR_BITS   9
>  #define BDRV_SECTOR_SIZE   (1ULL << BDRV_SECTOR_BITS)
>  
> +/*
> + * Get the first most significant bit of wp. If it is zero, then
> + * the zone type is SWR.
> + */
> +#define BDRV_ZT_IS_CONV(wp)    (wp & (1ULL << 63))
> +
>  #define BDRV_REQUEST_MAX_SECTORS MIN_CONST(SIZE_MAX >> BDRV_SECTOR_BITS, \
>                                             INT_MAX >> BDRV_SECTOR_BITS)
>  #define BDRV_REQUEST_MAX_BYTES (BDRV_REQUEST_MAX_SECTORS << BDRV_SECTOR_BITS)
> diff --git a/include/block/block_int-common.h b/include/block/block_int-
> common.h
> index 37dddc603c..e3136146aa 100644
> --- a/include/block/block_int-common.h
> +++ b/include/block/block_int-common.h
> @@ -857,6 +857,9 @@ typedef struct BlockLimits {
>  
>      /* device capacity expressed in bytes */
>      int64_t capacity;
> +
> +    /* array of write pointers' location of each zone in the zoned device. */
> +    BlockZoneWps *wps;
>  } BlockLimits;
>  
>  typedef struct BdrvOpBlocker BdrvOpBlocker;

Reply via email to