From: Gollu Appalanaidu <anaidu.gollu@samsung.com>

Add the Compare command.

This implementation uses a bounce buffer to read in the data from
storage and then compare with the host supplied buffer.

Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
[k.jensen: rebased]
Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
---
 hw/block/nvme.c       | 100 +++++++++++++++++++++++++++++++++++++++++-
 hw/block/trace-events |   2 +
 2 files changed, 101 insertions(+), 1 deletion(-)

diff --git a/hw/block/nvme.c b/hw/block/nvme.c
index f7f888402b06..f88710ca3948 100644
--- a/hw/block/nvme.c
+++ b/hw/block/nvme.c
@@ -999,6 +999,50 @@ static void nvme_aio_discard_cb(void *opaque, int ret)
     nvme_enqueue_req_completion(nvme_cq(req), req);
 }
 
+struct nvme_compare_ctx {
+    QEMUIOVector iov;
+    uint8_t *bounce;
+    size_t len;
+};
+
+static void nvme_compare_cb(void *opaque, int ret)
+{
+    NvmeRequest *req = opaque;
+    NvmeNamespace *ns = req->ns;
+    struct nvme_compare_ctx *ctx = req->opaque;
+    g_autofree uint8_t *buf = NULL;
+    uint16_t status;
+
+    trace_pci_nvme_compare_cb(nvme_cid(req));
+
+    if (!ret) {
+        block_acct_done(blk_get_stats(ns->blkconf.blk), &req->acct);
+    } else {
+        block_acct_failed(blk_get_stats(ns->blkconf.blk), &req->acct);
+        nvme_aio_err(req, ret);
+        goto out;
+    }
+
+    buf = g_malloc(ctx->len);
+
+    status = nvme_dma(nvme_ctrl(req), buf, ctx->len, DMA_DIRECTION_TO_DEVICE,
+                      req);
+    if (status) {
+        goto out;
+    }
+
+    if (memcmp(buf, ctx->bounce, ctx->len)) {
+        req->status = NVME_CMP_FAILURE;
+    }
+
+out:
+    qemu_iovec_destroy(&ctx->iov);
+    g_free(ctx->bounce);
+    g_free(ctx);
+
+    nvme_enqueue_req_completion(nvme_cq(req), req);
+}
+
 static uint16_t nvme_dsm(NvmeCtrl *n, NvmeRequest *req)
 {
     NvmeNamespace *ns = req->ns;
@@ -1072,6 +1116,57 @@ static uint16_t nvme_dsm(NvmeCtrl *n, NvmeRequest *req)
     return status;
 }
 
+static uint16_t nvme_compare(NvmeCtrl *n, NvmeRequest *req)
+{
+    NvmeRwCmd *rw = (NvmeRwCmd *)&req->cmd;
+    NvmeNamespace *ns = req->ns;
+    BlockBackend *blk = ns->blkconf.blk;
+    uint64_t slba = le64_to_cpu(rw->slba);
+    uint32_t nlb = le16_to_cpu(rw->nlb) + 1;
+    size_t len = nvme_l2b(ns, nlb);
+    int64_t offset = nvme_l2b(ns, slba);
+    uint8_t *bounce = NULL;
+    struct nvme_compare_ctx *ctx = NULL;
+    uint16_t status;
+
+    trace_pci_nvme_compare(nvme_cid(req), nvme_nsid(ns), slba, nlb);
+
+    status = nvme_check_mdts(n, len);
+    if (status) {
+        trace_pci_nvme_err_mdts(nvme_cid(req), len);
+        return status;
+    }
+
+    status = nvme_check_bounds(ns, slba, nlb);
+    if (status) {
+        trace_pci_nvme_err_invalid_lba_range(slba, nlb, ns->id_ns.nsze);
+        return status;
+    }
+
+    if (NVME_ERR_REC_DULBE(ns->features.err_rec)) {
+        status = nvme_check_dulbe(ns, slba, nlb);
+        if (status) {
+            return status;
+        }
+    }
+
+    bounce = g_malloc(len);
+
+    ctx = g_new(struct nvme_compare_ctx, 1);
+    ctx->bounce = bounce;
+    ctx->len = len;
+
+    req->opaque = ctx;
+
+    qemu_iovec_init(&ctx->iov, 1);
+    qemu_iovec_add(&ctx->iov, bounce, len);
+
+    block_acct_start(blk_get_stats(blk), &req->acct, len, BLOCK_ACCT_READ);
+    blk_aio_preadv(blk, offset, &ctx->iov, 0, nvme_compare_cb, req);
+
+    return NVME_NO_COMPLETE;
+}
+
 static uint16_t nvme_flush(NvmeCtrl *n, NvmeRequest *req)
 {
     block_acct_start(blk_get_stats(req->ns->blkconf.blk), &req->acct, 0,
@@ -1201,6 +1296,8 @@ static uint16_t nvme_io_cmd(NvmeCtrl *n, NvmeRequest *req)
     case NVME_CMD_WRITE:
     case NVME_CMD_READ:
         return nvme_rw(n, req);
+    case NVME_CMD_COMPARE:
+        return nvme_compare(n, req);
     case NVME_CMD_DSM:
         return nvme_dsm(n, req);
     default:
@@ -2927,7 +3024,8 @@ static void nvme_init_ctrl(NvmeCtrl *n, PCIDevice *pci_dev)
     id->cqes = (0x4 << 4) | 0x4;
     id->nn = cpu_to_le32(n->num_namespaces);
     id->oncs = cpu_to_le16(NVME_ONCS_WRITE_ZEROES | NVME_ONCS_TIMESTAMP |
-                           NVME_ONCS_FEATURES | NVME_ONCS_DSM);
+                           NVME_ONCS_FEATURES | NVME_ONCS_DSM |
+                           NVME_ONCS_COMPARE);
 
     id->vwc = 0x1;
     id->sgls = cpu_to_le32(NVME_CTRL_SGLS_SUPPORT_NO_ALIGN |
diff --git a/hw/block/trace-events b/hw/block/trace-events
index 1ffe0b3f76b5..68a4c8ed35e0 100644
--- a/hw/block/trace-events
+++ b/hw/block/trace-events
@@ -46,6 +46,8 @@ pci_nvme_write_zeroes(uint16_t cid, uint32_t nsid, uint64_t slba, uint32_t nlb)
 pci_nvme_block_status(int64_t offset, int64_t bytes, int64_t pnum, int ret, bool zeroed) "offset %"PRId64" bytes %"PRId64" pnum %"PRId64" ret 0x%x zeroed %d"
 pci_nvme_dsm(uint16_t cid, uint32_t nsid, uint32_t nr, uint32_t attr) "cid %"PRIu16" nsid %"PRIu32" nr %"PRIu32" attr 0x%"PRIx32""
 pci_nvme_dsm_deallocate(uint16_t cid, uint32_t nsid, uint64_t slba, uint32_t nlb) "cid %"PRIu16" nsid %"PRIu32" slba %"PRIu64" nlb %"PRIu32""
+pci_nvme_compare(uint16_t cid, uint32_t nsid, uint64_t slba, uint32_t nlb) "cid %"PRIu16" nsid %"PRIu32" slba 0x%"PRIx64" nlb %"PRIu32""
+pci_nvme_compare_cb(uint16_t cid) "cid %"PRIu16""
 pci_nvme_aio_discard_cb(uint16_t cid) "cid %"PRIu16""
 pci_nvme_create_sq(uint64_t addr, uint16_t sqid, uint16_t cqid, uint16_t qsize, uint16_t qflags) "create submission queue, addr=0x%"PRIx64", sqid=%"PRIu16", cqid=%"PRIu16", qsize=%"PRIu16", qflags=%"PRIu16""
 pci_nvme_create_cq(uint64_t addr, uint16_t cqid, uint16_t vector, uint16_t size, uint16_t qflags, int ien) "create completion queue, addr=0x%"PRIx64", cqid=%"PRIu16", vector=%"PRIu16", qsize=%"PRIu16", qflags=%"PRIu16", ien=%d"
-- 
2.29.2

Hello,

On 20-11-24 08:37:14, Klaus Jensen wrote:
> From: Gollu Appalanaidu <anaidu.gollu@samsung.com>
> 
> Add the Compare command.
> 
> This implementation uses a bounce buffer to read in the data from
> storage and then compare with the host supplied buffer.
> 
> Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
> [k.jensen: rebased]
> Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
> ---
>  hw/block/nvme.c       | 100 +++++++++++++++++++++++++++++++++++++++++-
>  hw/block/trace-events |   2 +
>  2 files changed, 101 insertions(+), 1 deletion(-)
> 
> diff --git a/hw/block/nvme.c b/hw/block/nvme.c
> index f7f888402b06..f88710ca3948 100644
> --- a/hw/block/nvme.c
> +++ b/hw/block/nvme.c
> @@ -999,6 +999,50 @@ static void nvme_aio_discard_cb(void *opaque, int ret)
>      nvme_enqueue_req_completion(nvme_cq(req), req);
>  }
>  
> +struct nvme_compare_ctx {
> +    QEMUIOVector iov;
> +    uint8_t *bounce;
> +    size_t len;
> +};
> +
> +static void nvme_compare_cb(void *opaque, int ret)
> +{
> +    NvmeRequest *req = opaque;
> +    NvmeNamespace *ns = req->ns;
> +    struct nvme_compare_ctx *ctx = req->opaque;
> +    g_autofree uint8_t *buf = NULL;

nit-picking here: unnecessary initialization to NULL.

> +    uint16_t status;
> +
> +    trace_pci_nvme_compare_cb(nvme_cid(req));
> +
> +    if (!ret) {
> +        block_acct_done(blk_get_stats(ns->blkconf.blk), &req->acct);
> +    } else {
> +        block_acct_failed(blk_get_stats(ns->blkconf.blk), &req->acct);
> +        nvme_aio_err(req, ret);
> +        goto out;
> +    }
> +
> +    buf = g_malloc(ctx->len);
> +
> +    status = nvme_dma(nvme_ctrl(req), buf, ctx->len, DMA_DIRECTION_TO_DEVICE,
> +                      req);
> +    if (status) {
> +        goto out;
> +    }

Don't we need to give status value to req->status in case of
(status != 0)?  If we don't give it to req->status, it will complete
with success status code even it fails during the nvme_dma().

On Nov 25 23:11, Minwoo Im wrote:
> Hello,
> 
> On 20-11-24 08:37:14, Klaus Jensen wrote:
> > From: Gollu Appalanaidu <anaidu.gollu@samsung.com>
> > 
> > Add the Compare command.
> > 
> > This implementation uses a bounce buffer to read in the data from
> > storage and then compare with the host supplied buffer.
> > 
> > Signed-off-by: Gollu Appalanaidu <anaidu.gollu@samsung.com>
> > [k.jensen: rebased]
> > Signed-off-by: Klaus Jensen <k.jensen@samsung.com>
> > ---
> >  hw/block/nvme.c       | 100 +++++++++++++++++++++++++++++++++++++++++-
> >  hw/block/trace-events |   2 +
> >  2 files changed, 101 insertions(+), 1 deletion(-)
> > 
> > diff --git a/hw/block/nvme.c b/hw/block/nvme.c
> > index f7f888402b06..f88710ca3948 100644
> > --- a/hw/block/nvme.c
> > +++ b/hw/block/nvme.c
> > @@ -999,6 +999,50 @@ static void nvme_aio_discard_cb(void *opaque, int ret)
> >      nvme_enqueue_req_completion(nvme_cq(req), req);
> >  }
> >  
> > +struct nvme_compare_ctx {
> > +    QEMUIOVector iov;
> > +    uint8_t *bounce;
> > +    size_t len;
> > +};
> > +
> > +static void nvme_compare_cb(void *opaque, int ret)
> > +{
> > +    NvmeRequest *req = opaque;
> > +    NvmeNamespace *ns = req->ns;
> > +    struct nvme_compare_ctx *ctx = req->opaque;
> > +    g_autofree uint8_t *buf = NULL;
> 
> nit-picking here: unnecessary initialization to NULL.
> 

I don't think it is unnecessary when it using g_autofree?

> > +    uint16_t status;
> > +
> > +    trace_pci_nvme_compare_cb(nvme_cid(req));
> > +
> > +    if (!ret) {
> > +        block_acct_done(blk_get_stats(ns->blkconf.blk), &req->acct);
> > +    } else {
> > +        block_acct_failed(blk_get_stats(ns->blkconf.blk), &req->acct);
> > +        nvme_aio_err(req, ret);
> > +        goto out;
> > +    }
> > +
> > +    buf = g_malloc(ctx->len);
> > +
> > +    status = nvme_dma(nvme_ctrl(req), buf, ctx->len, DMA_DIRECTION_TO_DEVICE,
> > +                      req);
> > +    if (status) {
> > +        goto out;
> > +    }
> 
> Don't we need to give status value to req->status in case of
> (status != 0)?  If we don't give it to req->status, it will complete
> with success status code even it fails during the nvme_dma().
> 

Nice catch! nvme_aio_err normally takes care of this for blk/aio errors,
but this one slipped. Thanks!