From: Thanos Makatos <thanos@nutanix.com>

This patch allows specifying an NVMe namespace ID in the boot ROM file
in the form /pci@0cf8/*@*/nvme/ns@*, where the 2nd component is the
PCI device and the last component is the namespace ID.

E.g. booting from namespace 2:

    boot order:
    1: /pci@i0cf8/pci4e58,0001@6/nvme/ns@2
    2: /rom@genroms/bootfail.bin
    3: HAL
    ...
    |3ffa9000| Searching bootorder for: /pci@i0cf8/*@6/nvme/ns@2
    |3ffa9000| NS 2 prio 1
    |3ffa9000| phys_alloc zone=0x3ffbfeb8 size=80 align=10 ret=3ffac6e0
    (detail=0x3ffac6b0)
    |3ffa9000| phys_alloc zone=0x3ffbfeb8 size=24 align=10 ret=3ffac690
    (detail=0x3ffac660)
    |3ffa9000| Registering bootable: NVMe NS 2: 19073 MiB (39062500 512-byte
    blocks + 0-byte metadata)
     (type:2 prio:1 data:f4050)
    |3ffa9000| NVMe initialization complete!
    \3ffa9000/ End thread

If we find a specific namespace we're looking for then we stop probing
all other namespaces as there's no point in doing so. If a namespace is
not specified in the boot ROM file then we default to the old way of
booting from NVMe.

As a future optimization we can avoid keeping information for all
namespaces and operate on a single namespace at at time. This way we
won't run out of memory even if the NVMe controller has the maximum
number of namespaces (2^32 - 1).

The corresponding QEMU patches haven't been upstreamed yet.

Signed-off-by: Thanos Makatos <thanos.makatos@nutanix.com>

--

Resending patch since it never reached the list, including a more detailed
commit message.

---
 src/boot.c    | 12 ++++++++++++
 src/hw/nvme.c | 49 +++++++++++++++++++++++++++++++++++++++----------
 src/util.h    |  1 +
 3 files changed, 52 insertions(+), 10 deletions(-)

diff --git a/src/boot.c b/src/boot.c
index 1effd80..6c2a150 100644
--- a/src/boot.c
+++ b/src/boot.c
@@ -325,6 +325,18 @@ int bootprio_find_mmio_device(void *mmio)
     return find_prio(desc);
 }
 
+int bootprio_find_nvme_device(struct pci_device *pci, u32 ns_id)
+{
+    if (CONFIG_CSM)
+        return csm_bootprio_pci(pci);
+    if (!CONFIG_BOOTORDER)
+        return -1;
+    char desc[256], *p;
+    p = build_pci_path(desc, sizeof(desc), "*", pci);
+    snprintf(p, desc+sizeof(desc)-p, "/nvme/ns@%u", ns_id);
+    return find_prio(desc);
+}
+
 int bootprio_find_scsi_device(struct pci_device *pci, int target, int lun)
 {
     if (!CONFIG_BOOTORDER)
diff --git a/src/hw/nvme.c b/src/hw/nvme.c
index f26b811..b35a273 100644
--- a/src/hw/nvme.c
+++ b/src/hw/nvme.c
@@ -233,10 +233,23 @@ nvme_admin_identify_ns(struct nvme_ctrl *ctrl, u32 ns_id)
                                 ns_id)->ns;
 }
 
-static void
+static char*
+nvme_ns_desc(const struct nvme_namespace *ns, u32 ns_id)
+{
+    return znprintf(MAXDESCSIZE, "NVMe NS %u: %llu MiB (%llu %u-byte "
+                                 "blocks + %u-byte metadata)\n",
+                                  ns_id, (ns->lba_count * ns->block_size) >> 20,
+                                  ns->lba_count, ns->block_size,
+                                  ns->metadata_size);
+}
+
+/* Returns 0 if the namespace is usable, -1 otherwise. */
+static int
 nvme_probe_ns(struct nvme_ctrl *ctrl, struct nvme_namespace *ns, u32 ns_id,
               u8 mdts)
 {
+    int ret = -1;
+
     ns->ctrl  = ctrl;
     ns->ns_id = ns_id;
 
@@ -288,16 +301,14 @@ nvme_probe_ns(struct nvme_ctrl *ctrl, struct nvme_namespace *ns, u32 ns_id,
 
     ns->dma_buffer = zalloc_page_aligned(&ZoneHigh, NVME_PAGE_SIZE);
 
-    char *desc = znprintf(MAXDESCSIZE, "NVMe NS %u: %llu MiB (%llu %u-byte "
-                          "blocks + %u-byte metadata)\n",
-                          ns_id, (ns->lba_count * ns->block_size) >> 20,
-                          ns->lba_count, ns->block_size, ns->metadata_size);
-
+    char *desc = nvme_ns_desc(ns, ns_id);
     dprintf(3, "%s", desc);
-    boot_add_hd(&ns->drive, desc, bootprio_find_pci_device(ctrl->pci));
+    free(desc);
+    ret = 0;
 
 free_buffer:
     free (id);
+    return ret;
 }
 
 
@@ -635,10 +646,28 @@ nvme_controller_enable(struct nvme_ctrl *ctrl)
     memset(ctrl->ns, 0, sizeof(*ctrl->ns) * ctrl->ns_count);
 
     /* Populate namespace IDs */
-    int ns_idx;
-    for (ns_idx = 0; ns_idx < ctrl->ns_count; ns_idx++) {
-        nvme_probe_ns(ctrl, &ctrl->ns[ns_idx], ns_idx + 1, identify->mdts);
+    int ns_idx, first_ns_idx = -1, prio = -1;
+    for (ns_idx = 0; ns_idx < ctrl->ns_count && prio == -1; ns_idx++) {
+        if (!nvme_probe_ns(ctrl, &ctrl->ns[ns_idx], ns_idx + 1, identify->mdts)) {
+            /* Check whether the NS is explicitly specified. */
+            if (first_ns_idx == -1) {
+                first_ns_idx = ns_idx;
+            }
+            prio = bootprio_find_nvme_device(ctrl->pci, ns_idx + 1);
+        }
     }
+     /* no NS specified, use the first one */
+    if (prio != -1) {
+        ns_idx--;
+    } else if (first_ns_idx != -1) {
+        ns_idx = first_ns_idx;
+        dprintf(3, "no NVMe NS specified, defaulting to NS %u\n", ns_idx + 1);
+        prio = bootprio_find_pci_device(ctrl->pci);
+    }
+    if (prio != -1) {
+        boot_add_hd(&ctrl->ns[ns_idx].drive,
+                    nvme_ns_desc(&ctrl->ns[ns_idx], ns_idx + 1), prio);
+   }
 
     dprintf(3, "NVMe initialization complete!\n");
     return 0;
diff --git a/src/util.h b/src/util.h
index 0de3522..ad2d0ef 100644
--- a/src/util.h
+++ b/src/util.h
@@ -31,6 +31,7 @@ u8 is_bootprio_strict(void);
 struct pci_device;
 int bootprio_find_pci_device(struct pci_device *pci);
 int bootprio_find_mmio_device(void *mmio);
+int bootprio_find_nvme_device(struct pci_device *pci, u32 ns_id);
 int bootprio_find_scsi_device(struct pci_device *pci, int target, int lun);
 int bootprio_find_scsi_mmio_device(void *mmio, int target, int lun);
 int bootprio_find_ata_device(struct pci_device *pci, int chanid, int slave);
-- 
2.22.3

_______________________________________________
SeaBIOS mailing list -- seabios@seabios.org
To unsubscribe send an email to seabios-leave@seabios.org