From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([208.118.235.92]:48383) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1USk8x-0006A1-7N for qemu-devel@nongnu.org; Thu, 18 Apr 2013 04:24:36 -0400 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1USk8v-0003Dp-MH for qemu-devel@nongnu.org; Thu, 18 Apr 2013 04:24:35 -0400 Received: from mail-da0-x22d.google.com ([2607:f8b0:400e:c00::22d]:46278) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1USk8v-0003DR-Cm for qemu-devel@nongnu.org; Thu, 18 Apr 2013 04:24:33 -0400 Received: by mail-da0-f45.google.com with SMTP id v40so1225839dad.18 for ; Thu, 18 Apr 2013 01:24:32 -0700 (PDT) From: Liu Yuan Date: Thu, 18 Apr 2013 16:24:22 +0800 Message-Id: <1366273462-16005-1-git-send-email-namei.unix@gmail.com> In-Reply-To: <1366263861-30146-1-git-send-email-namei.unix@gmail.com> References: <1366263861-30146-1-git-send-email-namei.unix@gmail.com> Subject: [Qemu-devel] [PATCH v8] sheepdog: add discard/trim support for sheepdog List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel@nongnu.org Cc: Kevin Wolf , Paolo Bonzini , sheepdog@lists.wpkg.org, Stefan Hajnoczi , MORITA Kazutaka From: Liu Yuan The 'TRIM' command from VM that is to release underlying data storage for better thin-provision is already supported by the Sheepdog. This patch adds the TRIM support at QEMU part. For older Sheepdog that doesn't support it, we return 0(success) to upper layer. Cc: MORITA Kazutaka Cc: Kevin Wolf Cc: Stefan Hajnoczi Cc: Paolo Bonzini Signed-off-by: Liu Yuan --- v8: - remove extra '\n' at error_report - add daemon spec in the error message - use per state bool indicator for discard operation v7: - add braces for 'if' to mute checkpath.pl v6: - add boolean indicator to disable discard operation when not supported by the server v5: - adjust macro numbering v4: - adjust discard macro - return success when operation is not supported by sheep - add coroutine_fn marker v3: - fix a silly accidental deletion of 'default' in switch clause. v2: - skip the object when it is not allocated block/sheepdog.c | 65 +++++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 64 insertions(+), 1 deletion(-) diff --git a/block/sheepdog.c b/block/sheepdog.c index 1c5b532..c099117 100644 --- a/block/sheepdog.c +++ b/block/sheepdog.c @@ -27,6 +27,8 @@ #define SD_OP_CREATE_AND_WRITE_OBJ 0x01 #define SD_OP_READ_OBJ 0x02 #define SD_OP_WRITE_OBJ 0x03 +/* 0x04 is used internally by Sheepdog */ +#define SD_OP_DISCARD_OBJ 0x05 #define SD_OP_NEW_VDI 0x11 #define SD_OP_LOCK_VDI 0x12 @@ -184,6 +186,7 @@ typedef struct SheepdogInode { uint32_t data_vdi_id[MAX_DATA_OBJS]; } SheepdogInode; + /* * 64 bit FNV-1a non-zero initial basis */ @@ -269,6 +272,7 @@ enum AIOCBState { AIOCB_WRITE_UDATA, AIOCB_READ_UDATA, AIOCB_FLUSH_CACHE, + AIOCB_DISCARD_OBJ, }; struct SheepdogAIOCB { @@ -298,6 +302,7 @@ typedef struct BDRVSheepdogState { char name[SD_MAX_VDI_LEN]; bool is_snapshot; uint32_t cache_flags; + bool discard_supported; char *host_spec; bool is_unix; @@ -656,7 +661,7 @@ static void coroutine_fn aio_read_response(void *opaque) int ret; AIOReq *aio_req = NULL; SheepdogAIOCB *acb; - unsigned long idx; + uint64_t idx; if (QLIST_EMPTY(&s->inflight_aio_head)) { goto out; @@ -727,6 +732,21 @@ static void coroutine_fn aio_read_response(void *opaque) rsp.result = SD_RES_SUCCESS; } break; + case AIOCB_DISCARD_OBJ: + switch (rsp.result) { + case SD_RES_INVALID_PARMS: + error_report("sheep(%s) doesn't support discard command", + s->host_spec); + rsp.result = SD_RES_SUCCESS; + s->discard_supported = false; + break; + case SD_RES_SUCCESS: + idx = data_oid_to_idx(aio_req->oid); + s->inode.data_vdi_id[idx] = 0; + break; + default: + break; + } } if (rsp.result != SD_RES_SUCCESS) { @@ -1016,6 +1036,9 @@ static int coroutine_fn add_aio_request(BDRVSheepdogState *s, AIOReq *aio_req, wlen = datalen; hdr.flags = SD_FLAG_CMD_WRITE | flags; break; + case AIOCB_DISCARD_OBJ: + hdr.opcode = SD_OP_DISCARD_OBJ; + break; } if (s->cache_flags) { @@ -1170,6 +1193,7 @@ static int sd_open(BlockDriverState *bs, const char *filename, if (flags & BDRV_O_NOCACHE) { s->cache_flags = SD_FLAG_CMD_DIRECT; } + s->discard_supported = true; if (snapid || tag[0] != '\0') { dprintf("%" PRIx32 " snapshot inode was open.\n", vid); @@ -1633,6 +1657,15 @@ static int coroutine_fn sd_co_rw_vector(void *p) flags = SD_FLAG_CMD_COW; } break; + case AIOCB_DISCARD_OBJ: + /* + * We discard the object only when the whole object is + * 1) allocated 2) trimmed. Otherwise, simply skip it. + */ + if (len != SD_DATA_OBJ_SIZE || inode->data_vdi_id[idx] == 0) { + goto done; + } + break; default: break; } @@ -2078,6 +2111,33 @@ static int sd_load_vmstate(BlockDriverState *bs, uint8_t *data, } +static coroutine_fn int sd_co_discard(BlockDriverState *bs, int64_t sector_num, + int nb_sectors) +{ + SheepdogAIOCB *acb; + QEMUIOVector dummy; + BDRVSheepdogState *s = bs->opaque; + int ret; + + if (!s->discard_supported) { + return 0; + } + + acb = sd_aio_setup(bs, &dummy, sector_num, nb_sectors); + acb->aiocb_type = AIOCB_DISCARD_OBJ; + acb->aio_done_func = sd_finish_aiocb; + + ret = sd_co_rw_vector(acb); + if (ret <= 0) { + qemu_aio_release(acb); + return ret; + } + + qemu_coroutine_yield(); + + return acb->ret; +} + static QEMUOptionParameter sd_create_options[] = { { .name = BLOCK_OPT_SIZE, @@ -2110,6 +2170,7 @@ static BlockDriver bdrv_sheepdog = { .bdrv_co_readv = sd_co_readv, .bdrv_co_writev = sd_co_writev, .bdrv_co_flush_to_disk = sd_co_flush_to_disk, + .bdrv_co_discard = sd_co_discard, .bdrv_snapshot_create = sd_snapshot_create, .bdrv_snapshot_goto = sd_snapshot_goto, @@ -2135,6 +2196,7 @@ static BlockDriver bdrv_sheepdog_tcp = { .bdrv_co_readv = sd_co_readv, .bdrv_co_writev = sd_co_writev, .bdrv_co_flush_to_disk = sd_co_flush_to_disk, + .bdrv_co_discard = sd_co_discard, .bdrv_snapshot_create = sd_snapshot_create, .bdrv_snapshot_goto = sd_snapshot_goto, @@ -2160,6 +2222,7 @@ static BlockDriver bdrv_sheepdog_unix = { .bdrv_co_readv = sd_co_readv, .bdrv_co_writev = sd_co_writev, .bdrv_co_flush_to_disk = sd_co_flush_to_disk, + .bdrv_co_discard = sd_co_discard, .bdrv_snapshot_create = sd_snapshot_create, .bdrv_snapshot_goto = sd_snapshot_goto, -- 1.7.9.5