[PATCH RESEND v8] Add cli and ioctl to forget scanned device(s)

[PATCH RESEND v8] Add cli and ioctl to forget scanned device(s)

[Anand Jain]

[applies on misc-next and for-next-20180801]

v8:
 Change log update in the kernel patch.

v7:
 Use struct btrfs_ioctl_vol_args (instead of struct
  btrfs_ioctl_vol_args_v2) as its inline with other ioctl
  btrfs-control
 The CLI usage/features remains same. However internally the ioctl flag
  is not required to delete all the unmounted devices. Instead leave
  btrfs_ioctl_vol_args::name NULL.

v6:
 Use the changed fn name btrfs_free_stale_devices().

 Change in title:
 Old v5:
 Cover-letter:
  [PATCH v5] Add cli and ioctl to ignore a scanned device
 Kernel:
  [PATCH v5] btrfs: introduce feature to ignore a btrfs device
 Progs:
  [PATCH v5] btrfs-progs: add 'btrfs device ignore' cli

v5:
  Adds feature to delete all stale devices
  Reuses btrfs_free_stale_devices() fn and so depends on the
    patch-set [1] in the ML.
  Uses struct btrfs_ioctl_vol_args_v2 instead of
    struct btrfs_ioctl_vol_args as arg
  Does the device path matching instead of btrfs_device matching
    (we won't delete the mounted device as btrfs_free_stale_devices()
    checks for it)
v4:
  No change. But as the ML thread may be confusing, so resend.
v3:
  No change. Send to correct ML.
v2:
  Accepts review from Nikolay, details are in the specific patch.
  Patch 1/2 is renamed from
    [PATCH 1/2] btrfs: refactor btrfs_free_stale_device() to get device list delete
  to
    [PATCH 1/2] btrfs: add function to device list delete

Adds cli and ioctl to forget a scanned device or forget all stale
devices in the kernel.

Anand Jain (1):
  btrfs: introduce feature to forget a btrfs device

 fs/btrfs/super.c           | 3 +++
 fs/btrfs/volumes.c         | 9 +++++++++
 fs/btrfs/volumes.h         | 1 +
 include/uapi/linux/btrfs.h | 2 ++
 4 files changed, 15 insertions(+)

Anand Jain (1):
  btrfs-progs: add cli to forget one or all scanned devices

 cmds-device.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 ioctl.h       |  2 ++
 2 files changed, 60 insertions(+)

-- 
2.7.0

[PATCH] btrfs: introduce feature to forget a btrfs device

[Anand Jain]

Support for a new command 'btrfs dev forget [dev]' is proposed here
to undo the effects of 'btrfs dev scan [dev]'. For this purpose
this patch proposes to use ioctl #5 as it was empty.
	IOW(BTRFS_IOCTL_MAGIC, 5, ..)
This patch adds new ioctl BTRFS_IOC_FORGET_DEV which can be sent from
the /dev/btrfs-control to forget one or all devices, (devices which are
not mounted) from the btrfs kernel.

The argument it takes is struct btrfs_ioctl_vol_args, and ::name can be
set to specify the device path. And all unmounted devices can be removed
from the kernel if no device path is provided.

Again, the devices are removed only if the relevant fsid aren't mounted.

This new cli can provide..
 . Release of unwanted btrfs_fs_devices and btrfs_devices memory if the
   device is not going to be mounted.
 . Ability to mount the device in degraded mode when one of the other
   device is corrupted like in split brain raid1.
 . Running test cases which requires btrfs.ko-reload if the rootfs
   is btrfs.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
---
 fs/btrfs/super.c           | 3 +++
 fs/btrfs/volumes.c         | 9 +++++++++
 fs/btrfs/volumes.h         | 1 +
 include/uapi/linux/btrfs.h | 2 ++
 4 files changed, 15 insertions(+)

diff --git a/fs/btrfs/super.c b/fs/btrfs/super.c
index 67de3c0fc85b..470a32af474e 100644
--- a/fs/btrfs/super.c
+++ b/fs/btrfs/super.c
@@ -2244,6 +2244,9 @@ static long btrfs_control_ioctl(struct file *file, unsigned int cmd,
 		ret = PTR_ERR_OR_ZERO(device);
 		mutex_unlock(&uuid_mutex);
 		break;
+	case BTRFS_IOC_FORGET_DEV:
+		ret = btrfs_forget_devices(vol->name);
+		break;
 	case BTRFS_IOC_DEVICES_READY:
 		mutex_lock(&uuid_mutex);
 		device = btrfs_scan_one_device(vol->name, FMODE_READ,
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 8844904f9009..cd54a926141a 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -1208,6 +1208,15 @@ static int btrfs_read_disk_super(struct block_device *bdev, u64 bytenr,
 	return 0;
 }
 
+int btrfs_forget_devices(const char *path)
+{
+	mutex_lock(&uuid_mutex);
+	btrfs_free_stale_devices(strlen(path) ? path:NULL, NULL);
+	mutex_unlock(&uuid_mutex);
+
+	return 0;
+}
+
 /*
  * Look for a btrfs signature on a device. This may be called out of the mount path
  * and we are not allowed to call set_blocksize during the scan. The superblock
diff --git a/fs/btrfs/volumes.h b/fs/btrfs/volumes.h
index 049619176831..1602b5faa7e7 100644
--- a/fs/btrfs/volumes.h
+++ b/fs/btrfs/volumes.h
@@ -405,6 +405,7 @@ int btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		       fmode_t flags, void *holder);
 struct btrfs_device *btrfs_scan_one_device(const char *path,
 					   fmode_t flags, void *holder);
+int btrfs_forget_devices(const char *path);
 int btrfs_close_devices(struct btrfs_fs_devices *fs_devices);
 void btrfs_free_extra_devids(struct btrfs_fs_devices *fs_devices, int step);
 void btrfs_assign_next_active_device(struct btrfs_device *device,
diff --git a/include/uapi/linux/btrfs.h b/include/uapi/linux/btrfs.h
index 5ca1d21fc4a7..b1be7f828cb4 100644
--- a/include/uapi/linux/btrfs.h
+++ b/include/uapi/linux/btrfs.h
@@ -836,6 +836,8 @@ enum btrfs_err_code {
 				   struct btrfs_ioctl_vol_args)
 #define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \
 				   struct btrfs_ioctl_vol_args)
+#define BTRFS_IOC_FORGET_DEV _IOW(BTRFS_IOCTL_MAGIC, 5, \
+				   struct btrfs_ioctl_vol_args)
 /* trans start and trans end are dangerous, and only for
  * use by applications that know how to avoid the
  * resulting deadlocks
-- 
2.7.0

[PATCH] btrfs-progs: add cli to forget one or all scanned devices

[Anand Jain]

This patch adds cli
  btrfs device forget [dev]
to remove the given device structure in the kernel if the device
is unmounted. If no argument is given it shall remove all stale
(device which are not mounted) from the kernel.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
---
 cmds-device.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 ioctl.h       |  2 ++
 2 files changed, 60 insertions(+)

diff --git a/cmds-device.c b/cmds-device.c
index 86459d1b9564..49cfd4b41adb 100644
--- a/cmds-device.c
+++ b/cmds-device.c
@@ -326,6 +326,63 @@ out:
 	return !!ret;
 }
 
+static const char * const cmd_device_forget_usage[] = {
+	"btrfs device forget [<device>]",
+	"Forget a stale device or all stale devices in btrfs.ko",
+	NULL
+};
+
+static int btrfs_forget_devices(char *path)
+{
+	struct btrfs_ioctl_vol_args args;
+	int ret;
+	int fd;
+
+	fd = open("/dev/btrfs-control", O_RDWR);
+	if (fd < 0)
+		return -errno;
+
+	memset(&args, 0, sizeof(args));
+	if (path)
+		strncpy_null(args.name, path);
+	ret = ioctl(fd, BTRFS_IOC_FORGET_DEV, &args);
+	if (ret)
+		ret = -errno;
+	close(fd);
+	return ret;
+}
+
+static int cmd_device_forget(int argc, char **argv)
+{
+	char *path;
+	int ret = 0;
+
+	if (check_argc_max(argc - optind, 1))
+		usage(cmd_device_forget_usage);
+
+	if (argc == 1) {
+		ret = btrfs_forget_devices(NULL);
+		if (ret)
+			error("Can't forget: %s", strerror(-ret));
+		return ret;
+	}
+
+	path = canonicalize_path(argv[1]);
+	if (!path) {
+		error("Could not canonicalize path '%s': %s",
+					argv[1], strerror(errno));
+		return -ENOENT;
+	}
+
+	ret  = btrfs_forget_devices(path);
+	if (ret)
+		error("Can't forget '%s': %s", path, strerror(-ret));
+
+	free(path);
+
+	return ret;
+}
+
 static const char * const cmd_device_ready_usage[] = {
 	"btrfs device ready <device>",
 	"Check device to see if it has all of its devices in cache for mounting",
@@ -601,6 +658,7 @@ const struct cmd_group device_cmd_group = {
 			CMD_ALIAS },
 		{ "remove", cmd_device_remove, cmd_device_remove_usage, NULL, 0 },
 		{ "scan", cmd_device_scan, cmd_device_scan_usage, NULL, 0 },
+		{ "forget", cmd_device_forget, cmd_device_forget_usage, NULL, 0 },
 		{ "ready", cmd_device_ready, cmd_device_ready_usage, NULL, 0 },
 		{ "stats", cmd_device_stats, cmd_device_stats_usage, NULL, 0 },
 		{ "usage", cmd_device_usage,
diff --git a/ioctl.h b/ioctl.h
index 709e996f401c..e27d80e09392 100644
--- a/ioctl.h
+++ b/ioctl.h
@@ -721,6 +721,8 @@ static inline char *btrfs_err_str(enum btrfs_err_code err_code)
 				   struct btrfs_ioctl_vol_args)
 #define BTRFS_IOC_SCAN_DEV _IOW(BTRFS_IOCTL_MAGIC, 4, \
 				   struct btrfs_ioctl_vol_args)
+#define BTRFS_IOC_FORGET_DEV _IOW(BTRFS_IOCTL_MAGIC, 5, \
+				   struct btrfs_ioctl_vol_args)
 /* trans start and trans end are dangerous, and only for
  * use by applications that know how to avoid the
  * resulting deadlocks
-- 
2.7.0

Re: [PATCH RESEND v8] Add cli and ioctl to forget scanned device(s)

[David Sterba]

On Mon, Aug 06, 2018 at 09:09:47AM +0800, Anand Jain wrote:
> Adds cli and ioctl to forget a scanned device or forget all stale
> devices in the kernel.

Please provide more details about your idea of the usecase, ie. how
excactly and when the command is supposed to be used.

I vaguely recollect that we've discussed the commandline interface but
don't remember the result. Nevertheless, I think that the 'forget'
command should be an option of 'device scan'.

  btrfs device scan --forget /dev/sda

Next I'm not sure the freeing all stale devices is a good idea. Should
it be more fine grained? Suppose there are several multi-device
filesystems on the host and some of them not mounted. The devices have
been scanned eg. via udev and the filesystems are ready to be mounted.
Calling 'forget all' will now prevent mount without another 'device
scan'.

What if a particular filesystem needs to forget the scanned devices, or
just one filesystem:device.

And there's the question how to specify the devices, it can be by device
path or uuid or maybe device id.

Re: [PATCH RESEND v8] Add cli and ioctl to forget scanned device(s)

[Anand Jain]

On 08/08/2018 01:51 AM, David Sterba wrote:
> On Mon, Aug 06, 2018 at 09:09:47AM +0800, Anand Jain wrote:
>> Adds cli and ioctl to forget a scanned device or forget all stale
>> devices in the kernel.
> 
> Please provide more details about your idea of the usecase, ie. how
> excactly and when the command is supposed to be used.

  Much of this details are in the kernel change log so I didn't add
  here, wonder if you miss that?

> I vaguely recollect that we've discussed the commandline interface but
> don't remember the result. Nevertheless, I think that the 'forget'
> command should be an option of 'device scan'.

  Right. Its here[1]. As I mentioned it had typo ?

  [1]
   https://patchwork.kernel.org/patch/10092511/

>    btrfs device scan --forget /dev/sda

  This syntax is is not in line with our other cli.
  Like: btrfs device add; btrfs device delete;

> Next I'm not sure the freeing all stale devices is a good idea.

   Its not a bad idea though. You suggested it here [1] (above).

< Should
> it be more fine grained?

  The current proposed cli is also fine grained.
     btrfs device forget [device-path]
  Where it checks and releases only [device-path]

> Suppose there are several multi-device
> filesystems on the host and some of them not mounted. The devices have
> been scanned eg. via udev and the filesystems are ready to be mounted.
> Calling 'forget all' will now prevent mount without another 'device
> scan'.

   Right. So if users intention is to mount then he should not call
   btrfs device forget. (with or without any device path).

> What if a particular filesystem needs to forget the scanned devices, or
> just one filesystem:device.

  The current cli in this patch supports this..
     btrfs device forget [device-path]

  So user can forget a device.

> And there's the question how to specify the devices, it can be by device
> path or uuid or maybe device id.

  The forget part is synonym to its 'btrfs device scan [device-path]'.

  The stale devices can be of any type. As of now there is no way user
  can know their stale devices. Only way they can clean them all is
  by running
    btrfs device forget <-- without any option so cleans all unmounted
   device.

  So providing the fsid/uuid option is of no use.

Thanks, Anand

> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [PATCH RESEND v8] Add cli and ioctl to forget scanned device(s)

[Anand Jain]

Hi David,

  Just a gentle ping..

Thanks, Anand


On 08/08/2018 07:33 AM, Anand Jain wrote:
> 
> 
> On 08/08/2018 01:51 AM, David Sterba wrote:
>> On Mon, Aug 06, 2018 at 09:09:47AM +0800, Anand Jain wrote:
>>> Adds cli and ioctl to forget a scanned device or forget all stale
>>> devices in the kernel.
>>
>> Please provide more details about your idea of the usecase, ie. how
>> excactly and when the command is supposed to be used.
> 
>   Much of this details are in the kernel change log so I didn't add
>   here, wonder if you miss that?
> 
>> I vaguely recollect that we've discussed the commandline interface but
>> don't remember the result. Nevertheless, I think that the 'forget'
>> command should be an option of 'device scan'.
> 
>   Right. Its here[1]. As I mentioned it had typo ?
> 
>   [1]
>    https://patchwork.kernel.org/patch/10092511/
> 
>>    btrfs device scan --forget /dev/sda
> 
>   This syntax is is not in line with our other cli.
>   Like: btrfs device add; btrfs device delete;
> 
>> Next I'm not sure the freeing all stale devices is a good idea.
> 
>    Its not a bad idea though. You suggested it here [1] (above).
> 
> < Should
>> it be more fine grained?
> 
>   The current proposed cli is also fine grained.
>      btrfs device forget [device-path]
>   Where it checks and releases only [device-path]
> 
>> Suppose there are several multi-device
>> filesystems on the host and some of them not mounted. The devices have
>> been scanned eg. via udev and the filesystems are ready to be mounted.
>> Calling 'forget all' will now prevent mount without another 'device
>> scan'.
> 
>    Right. So if users intention is to mount then he should not call
>    btrfs device forget. (with or without any device path).
> 
>> What if a particular filesystem needs to forget the scanned devices, or
>> just one filesystem:device.
> 
>   The current cli in this patch supports this..
>      btrfs device forget [device-path]
> 
>   So user can forget a device.
> 
>> And there's the question how to specify the devices, it can be by device
>> path or uuid or maybe device id.
> 
>   The forget part is synonym to its 'btrfs device scan [device-path]'.
> 
>   The stale devices can be of any type. As of now there is no way user
>   can know their stale devices. Only way they can clean them all is
>   by running
>     btrfs device forget <-- without any option so cleans all unmounted
>    device.
> 
>   So providing the fsid/uuid option is of no use.
> 
> Thanks, Anand
> 
>> -- 
>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>