Hi Christoph, Mike and Hannes
On 2017/10/29 18:01, Hannes Reinecke wrote:
> After all, Linux is about choice, not about forcing users to do things
> in one way only.
I have added an option CONFIG_NVME_SHOW_CTRL_BLK_DEV to show per-controller
block devices nodes.
It is tested simply that dm-multipath and nvme-mpath can coexist.
This patch, based on the Christoph V5 of nvme-mpath, is in the following.
Does it look good for you ?
Regards
Guan
--
>From de3f446af6591d68ef84333138e744f12db4d695 Mon Sep 17 00:00:00 2001
From: Junxiong Guan <guanjunxiong@xxxxxxxxxx>
Date: Mon, 30 Oct 2017 04:59:20 -0400
Subject: [PATCH] nvme: add an option to show per-controller block devices
nodes
Signed-off-by: Junxiong Guan <guanjunxiong@xxxxxxxxxx>
---
drivers/nvme/host/Kconfig | 9 ++++++
drivers/nvme/host/core.c | 80 +++++++++++++++++++++++++++++++++++++++++++++--
2 files changed, 87 insertions(+), 2 deletions(-)
diff --git a/drivers/nvme/host/Kconfig b/drivers/nvme/host/Kconfig
index 46d6cb1e03bd..725bff035f38 100644
--- a/drivers/nvme/host/Kconfig
+++ b/drivers/nvme/host/Kconfig
@@ -13,6 +13,15 @@ config BLK_DEV_NVME
To compile this driver as a module, choose M here: the
module will be called nvme.
+config NVME_SHOW_CTRL_BLK_DEV
+ bool "Show per-controller block devices of NVMe"
+ depends on NVME_CORE
+ ---help---
+ This adds support to show per-controller block devices nodes in
+ the user space.
+
+ If unsure, say N.
+
config NVME_FABRICS
tristate
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index 334735db90c8..ae37e274108c 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -72,6 +72,9 @@ static DEFINE_IDA(nvme_subsystems_ida);
static LIST_HEAD(nvme_subsystems);
static DEFINE_MUTEX(nvme_subsystems_lock);
+#ifdef CONFIG_NVME_SHOW_CTRL_BLK_DEV
+static DEFINE_SPINLOCK(dev_list_lock);
+#endif
static DEFINE_IDA(nvme_instance_ida);
static dev_t nvme_chr_devt;
static struct class *nvme_class;
@@ -357,6 +360,14 @@ static void nvme_free_ns(struct kref *kref)
if (ns->ndev)
nvme_nvm_unregister(ns);
+
+#ifdef CONFIG_NVME_SHOW_CTRL_BLK_DEV
+ if (ns->disk) {
+ spin_lock(&dev_list_lock);
+ ns->disk->private_data = NULL;
+ spin_unlock(&dev_list_lock);
+ }
+#endif
put_disk(ns->disk);
nvme_put_ns_head(ns->head);
nvme_put_ctrl(ns->ctrl);
@@ -1127,10 +1138,32 @@ static int nvme_ns_ioctl(struct nvme_ns *ns, unsigned int cmd,
}
}
-/* should never be called due to GENHD_FL_HIDDEN */
static int nvme_open(struct block_device *bdev, fmode_t mode)
{
+#ifdef CONFIG_NVME_SHOW_CTRL_BLK_DEV
+ struct nvme_ns *ns;
+
+ spin_lock(&dev_list_lock);
+ ns = bdev->bd_disk->private_data;
+ if (ns) {
+ if (!kref_get_unless_zero(&ns->kref))
+ goto fail;
+ if (!try_module_get(ns->ctrl->ops->module))
+ goto fail_put_ns;
+ }
+ spin_unlock(&dev_list_lock);
+
+ return ns ? 0 : -ENXIO;
+
+fail_put_ns:
+ kref_put(&ns->kref, nvme_free_ns);
+fail:
+ spin_unlock(&dev_list_lock);
+ return -ENXIO;
+#else
+ /* should never be called due to GENHD_FL_HIDDEN */
return WARN_ON_ONCE(-ENXIO);
+#endif /* CONFIG_NVME_SHOW_CTRL_BLK_DEV */
}
static int nvme_getgeo(struct block_device *bdev, struct hd_geometry *geo)
@@ -1392,10 +1425,15 @@ static char nvme_pr_type(enum pr_type type)
static int nvme_pr_command(struct block_device *bdev, u32 cdw10,
u64 key, u64 sa_key, u8 op)
{
+#ifdef CONFIG_NVME_SHOW_CTRL_BLK_DEV
+ struct nvme_ns *ns = bdev->bd_disk->private_data;
+ struct nvme_ns_head *head = ns->head;
+#else
struct nvme_ns_head *head = bdev->bd_disk->private_data;
struct nvme_ns *ns;
- struct nvme_command c;
int srcu_idx, ret;
+#endif
+ struct nvme_command c;
u8 data[16] = { 0, };
put_unaligned_le64(key, &data[0]);
@@ -1406,6 +1444,9 @@ static int nvme_pr_command(struct block_device *bdev, u32 cdw10,
c.common.nsid = cpu_to_le32(head->ns_id);
c.common.cdw10[0] = cpu_to_le32(cdw10);
+#ifdef CONFIG_NVME_SHOW_CTRL_BLK_DEV
+ return nvme_submit_sync_cmd(ns->queue, &c, data, 16);
+#else
srcu_idx = srcu_read_lock(&head->srcu);
ns = nvme_find_path(head);
if (likely(ns))
@@ -1413,7 +1454,9 @@ static int nvme_pr_command(struct block_device *bdev, u32 cdw10,
else
ret = -EWOULDBLOCK;
srcu_read_unlock(&head->srcu, srcu_idx);
+
return ret;
+#endif
}
static int nvme_pr_register(struct block_device *bdev, u64 old,
@@ -1492,6 +1535,34 @@ int nvme_sec_submit(void *data, u16 spsp, u8 secp, void *buffer, size_t len,
EXPORT_SYMBOL_GPL(nvme_sec_submit);
#endif /* CONFIG_BLK_SED_OPAL */
+#ifdef CONFIG_NVME_SHOW_CTRL_BLK_DEV
+static void nvme_release(struct gendisk *disk, fmode_t mode)
+{
+ struct nvme_ns *ns = disk->private_data;
+
+ module_put(ns->ctrl->ops->module);
+ nvme_put_ns(ns);
+}
+
+static int nvme_ioctl(struct block_device *bdev, fmode_t mode,
+ unsigned int cmd, unsigned long arg)
+{
+ struct nvme_ns *ns = bdev->bd_disk->private_data;
+
+ return nvme_ns_ioctl(ns, cmd, arg);
+}
+
+static const struct block_device_operations nvme_fops = {
+ .owner = THIS_MODULE,
+ .ioctl = nvme_ioctl,
+ .compat_ioctl = nvme_ioctl,
+ .open = nvme_open,
+ .release = nvme_release,
+ .getgeo = nvme_getgeo,
+ .revalidate_disk= nvme_revalidate_disk,
+ .pr_ops =&nvme_pr_ops,
+};
+#else
/*
* While we don't expose the per-controller devices to userspace we still
* need valid file operations for them, for one because the block layer
@@ -1503,6 +1574,7 @@ static const struct block_device_operations nvme_fops = {
.open = nvme_open,
.revalidate_disk= nvme_revalidate_disk,
};
+#endif /* CONFIG_NVME_SHOW_CTRL_BLK_DEV */
static int nvme_wait_ready(struct nvme_ctrl *ctrl, u64 cap, bool enabled)
{
@@ -2875,7 +2947,11 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid)
disk->fops = &nvme_fops;
disk->private_data = ns;
disk->queue = ns->queue;
+#ifdef CONFIG_NVME_SHOW_CTRL_BLK_DEV
+ disk->flags = GENHD_FL_EXT_DEVT;
+#else
disk->flags = GENHD_FL_HIDDEN;
+#endif
memcpy(disk->disk_name, disk_name, DISK_NAME_LEN);
ns->disk = disk;
--
2.11.1
