Re: [Qemu-devel] [PATCH] ISCSI: Switch to using READ16/WRITE16 for I/O to the LUN.

[Paolo Bonzini]

Il 22/05/2012 12:10, Ronnie Sahlberg ha scritto:
> This allows using LUNs bigger than 2TB.

Applied to scsi-next for 1.2, thanks.

Paolo

> Signed-off-by: Ronnie Sahlberg <address@hidden>
> ---
>  block/iscsi.c |  101 ++++++++++++++++++++++++++++++++++++++++----------------
>  trace-events  |    4 +-
>  2 files changed, 74 insertions(+), 31 deletions(-)
> 
> diff --git a/block/iscsi.c b/block/iscsi.c
> index f956824..ed1ad7b 100644
> --- a/block/iscsi.c
> +++ b/block/iscsi.c
> @@ -25,6 +25,7 @@
>  #include "config-host.h"
>  
>  #include <poll.h>
> +#include <arpa/inet.h>
>  #include "qemu-common.h"
>  #include "qemu-error.h"
>  #include "block_int.h"
> @@ -168,12 +169,12 @@ iscsi_readv_writev_bh_cb(void *p)
>  
>  
>  static void
> -iscsi_aio_write10_cb(struct iscsi_context *iscsi, int status,
> +iscsi_aio_write16_cb(struct iscsi_context *iscsi, int status,
>                       void *command_data, void *opaque)
>  {
>      IscsiAIOCB *acb = opaque;
>  
> -    trace_iscsi_aio_write10_cb(iscsi, status, acb, acb->canceled);
> +    trace_iscsi_aio_write16_cb(iscsi, status, acb, acb->canceled);
>  
>      g_free(acb->buf);
>  
> @@ -186,7 +187,7 @@ iscsi_aio_write10_cb(struct iscsi_context *iscsi, int 
> status,
>  
>      acb->status = 0;
>      if (status < 0) {
> -        error_report("Failed to write10 data to iSCSI lun. %s",
> +        error_report("Failed to write16 data to iSCSI lun. %s",
>                       iscsi_get_error(iscsi));
>          acb->status = -EIO;
>      }
> @@ -211,12 +212,9 @@ iscsi_aio_writev(BlockDriverState *bs, int64_t 
> sector_num,
>      struct iscsi_context *iscsi = iscsilun->iscsi;
>      IscsiAIOCB *acb;
>      size_t size;
> -    int fua = 0;
> -
> -    /* set FUA on writes when cache mode is write through */
> -    if (!(bs->open_flags & BDRV_O_CACHE_WB)) {
> -        fua = 1;
> -    }
> +    uint32_t num_sectors;
> +    uint64_t lba;
> +    struct iscsi_data data;
>  
>      acb = qemu_aio_get(&iscsi_aio_pool, bs, cb, opaque);
>      trace_iscsi_aio_writev(iscsi, sector_num, nb_sectors, opaque, acb);
> @@ -226,18 +224,44 @@ iscsi_aio_writev(BlockDriverState *bs, int64_t 
> sector_num,
>  
>      acb->canceled   = 0;
>  
> -    /* XXX we should pass the iovec to write10 to avoid the extra copy */
> +    /* XXX we should pass the iovec to write16 to avoid the extra copy */
>      /* this will allow us to get rid of 'buf' completely */
>      size = nb_sectors * BDRV_SECTOR_SIZE;
>      acb->buf = g_malloc(size);
>      qemu_iovec_to_buffer(acb->qiov, acb->buf);
> -    acb->task = iscsi_write10_task(iscsi, iscsilun->lun, acb->buf, size,
> -                              sector_qemu2lun(sector_num, iscsilun),
> -                              fua, 0, iscsilun->block_size,
> -                              iscsi_aio_write10_cb, acb);
> +
> +
> +    acb->task = malloc(sizeof(struct scsi_task));
>      if (acb->task == NULL) {
> -        error_report("iSCSI: Failed to send write10 command. %s",
> -                     iscsi_get_error(iscsi));
> +        error_report("iSCSI: Failed to allocate task for scsi WRITE16 "
> +                     "command. %s", iscsi_get_error(iscsi));
> +        qemu_aio_release(acb);
> +        return NULL;
> +    }
> +    memset(acb->task, 0, sizeof(struct scsi_task));
> +
> +    acb->task->xfer_dir = SCSI_XFER_WRITE;
> +    acb->task->cdb_size = 16;
> +    acb->task->cdb[0] = 0x8a;
> +    if (!(bs->open_flags & BDRV_O_CACHE_WB)) {
> +        /* set FUA on writes when cache mode is write through */
> +        acb->task->cdb[1] |= 0x04;
> +    }
> +    lba = sector_qemu2lun(sector_num, iscsilun);
> +    *(uint32_t *)&acb->task->cdb[2]  = htonl(lba >> 32);
> +    *(uint32_t *)&acb->task->cdb[6]  = htonl(lba & 0xffffffff);
> +    num_sectors = size / iscsilun->block_size;
> +    *(uint32_t *)&acb->task->cdb[10] = htonl(num_sectors);
> +    acb->task->expxferlen = size;
> +
> +    data.data = acb->buf;
> +    data.size = size;
> +
> +    if (iscsi_scsi_command_async(iscsi, iscsilun->lun, acb->task,
> +                                 iscsi_aio_write16_cb,
> +                                 &data,
> +                                 acb) != 0) {
> +        scsi_free_scsi_task(acb->task);
>          g_free(acb->buf);
>          qemu_aio_release(acb);
>          return NULL;
> @@ -249,12 +273,12 @@ iscsi_aio_writev(BlockDriverState *bs, int64_t 
> sector_num,
>  }
>  
>  static void
> -iscsi_aio_read10_cb(struct iscsi_context *iscsi, int status,
> +iscsi_aio_read16_cb(struct iscsi_context *iscsi, int status,
>                      void *command_data, void *opaque)
>  {
>      IscsiAIOCB *acb = opaque;
>  
> -    trace_iscsi_aio_read10_cb(iscsi, status, acb, acb->canceled);
> +    trace_iscsi_aio_read16_cb(iscsi, status, acb, acb->canceled);
>  
>      if (acb->canceled != 0) {
>          qemu_aio_release(acb);
> @@ -265,7 +289,7 @@ iscsi_aio_read10_cb(struct iscsi_context *iscsi, int 
> status,
>  
>      acb->status = 0;
>      if (status != 0) {
> -        error_report("Failed to read10 data from iSCSI lun. %s",
> +        error_report("Failed to read16 data from iSCSI lun. %s",
>                       iscsi_get_error(iscsi));
>          acb->status = -EIO;
>      }
> @@ -284,8 +308,10 @@ iscsi_aio_readv(BlockDriverState *bs, int64_t sector_num,
>      IscsiLun *iscsilun = bs->opaque;
>      struct iscsi_context *iscsi = iscsilun->iscsi;
>      IscsiAIOCB *acb;
> -    size_t qemu_read_size, lun_read_size;
> +    size_t qemu_read_size;
>      int i;
> +    uint64_t lba;
> +    uint32_t num_sectors;
>  
>      qemu_read_size = BDRV_SECTOR_SIZE * (size_t)nb_sectors;
>  
> @@ -310,16 +336,33 @@ iscsi_aio_readv(BlockDriverState *bs, int64_t 
> sector_num,
>          acb->read_offset  = bdrv_offset % iscsilun->block_size;
>      }
>  
> -    lun_read_size  = (qemu_read_size + iscsilun->block_size
> -                     + acb->read_offset - 1)
> -                     / iscsilun->block_size * iscsilun->block_size;
> -    acb->task = iscsi_read10_task(iscsi, iscsilun->lun,
> -                             sector_qemu2lun(sector_num, iscsilun),
> -                             lun_read_size, iscsilun->block_size,
> -                             iscsi_aio_read10_cb, acb);
> +    num_sectors  = (qemu_read_size + iscsilun->block_size
> +                    + acb->read_offset - 1)
> +                    / iscsilun->block_size;
> +
> +    acb->task = malloc(sizeof(struct scsi_task));
>      if (acb->task == NULL) {
> -        error_report("iSCSI: Failed to send read10 command. %s",
> -                     iscsi_get_error(iscsi));
> +        error_report("iSCSI: Failed to allocate task for scsi READ16 "
> +                     "command. %s", iscsi_get_error(iscsi));
> +        qemu_aio_release(acb);
> +        return NULL;
> +    }
> +    memset(acb->task, 0, sizeof(struct scsi_task));
> +
> +    acb->task->xfer_dir = SCSI_XFER_READ;
> +    acb->task->cdb_size = 16;
> +    acb->task->cdb[0]  = 0x88;
> +    lba = sector_qemu2lun(sector_num, iscsilun);
> +    *(uint32_t *)&acb->task->cdb[2]  = htonl(lba >> 32);
> +    *(uint32_t *)&acb->task->cdb[6]  = htonl(lba & 0xffffffff);
> +    *(uint32_t *)&acb->task->cdb[10] = htonl(num_sectors);
> +    acb->task->expxferlen = qemu_read_size;
> +
> +    if (iscsi_scsi_command_async(iscsi, iscsilun->lun, acb->task,
> +                                 iscsi_aio_read16_cb,
> +                                 NULL,
> +                                 acb) != 0) {
> +        scsi_free_scsi_task(acb->task);
>          qemu_aio_release(acb);
>          return NULL;
>      }
> diff --git a/trace-events b/trace-events
> index 87cb96c..45c6bc1 100644
> --- a/trace-events
> +++ b/trace-events
> @@ -602,9 +602,9 @@ escc_kbd_command(int val) "Command %d"
>  escc_sunmouse_event(int dx, int dy, int buttons_state) "dx=%d dy=%d 
> buttons=%01x"
>  
>  # block/iscsi.c
> -iscsi_aio_write10_cb(void *iscsi, int status, void *acb, int canceled) 
> "iscsi %p status %d acb %p canceled %d"
> +iscsi_aio_write16_cb(void *iscsi, int status, void *acb, int canceled) 
> "iscsi %p status %d acb %p canceled %d"
>  iscsi_aio_writev(void *iscsi, int64_t sector_num, int nb_sectors, void 
> *opaque, void *acb) "iscsi %p sector_num %"PRId64" nb_sectors %d opaque %p 
> acb %p"
> -iscsi_aio_read10_cb(void *iscsi, int status, void *acb, int canceled) "iscsi 
> %p status %d acb %p canceled %d"
> +iscsi_aio_read16_cb(void *iscsi, int status, void *acb, int canceled) "iscsi 
> %p status %d acb %p canceled %d"
>  iscsi_aio_readv(void *iscsi, int64_t sector_num, int nb_sectors, void 
> *opaque, void *acb) "iscsi %p sector_num %"PRId64" nb_sectors %d opaque %p 
> acb %p"
>  
>  # hw/esp.c