[PATCH] IMSM: Add warning message when x8-type device is used

[PATCH] IMSM: Add warning message when x8-type device is used

[Pawel Baldysiak]

This patch adds the warning message when x8-type device
is used with IMSM metadata. x8 device is a special
NVMe drive - two of them on a single PCIe card.
This card could be a single point of failure for
RAID levels different than RAID0. x8 devices have
serial number ending with "-A/-B" or "-1/-2".

Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
Reviewed-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
---
 super-intel.c | 42 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 42 insertions(+)

diff --git a/super-intel.c b/super-intel.c
index 5c6ab05..e860e75 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -5139,9 +5139,51 @@ static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
 	rv = imsm_read_serial(fd, devname, dd->serial);
 	if (rv) {
 		pr_err("failed to retrieve scsi serial, aborting\n");
+		if (dd->devname)
+			free(dd->devname);
 		free(dd);
 		abort();
 	}
+	if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
+	   (super->hba->type == SYS_DEV_VMD))) {
+		int i;
+		char *devpath = diskfd_to_devpath(fd);
+		char controller_path[PATH_MAX];
+
+		if (!devpath) {
+			pr_err("failed to get devpath, aborting\n");
+			if (dd->devname)
+				free(dd->devname);
+			free(dd);
+			return 1;
+		}
+
+		snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
+		free(devpath);
+
+		if (devpath_to_vendor(controller_path) == 0x8086) {
+			/*
+			 * If Intel's NVMe drive has serial ended with
+			 * "-A","-B","-1" or "-2" it means that this is "x8"
+			 * device (double drive on single PCIe card).
+			 * User should be warned about potential data loss.
+			 */
+			for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
+				/* Skip empty character at the end */
+				if (dd->serial[i] == 0)
+					continue;
+
+				if (((dd->serial[i] == 'A') || (dd->serial[i] == 'B') ||
+				   (dd->serial[i] == '1') || (dd->serial[i] == '2')) &&
+				   (dd->serial[i-1] == '-'))
+					pr_err("\tThe action you are about to take may put your data at risk.\n"
+						"\tPlease note that x8 devices may consist of two separate x4 devices "
+						"located on a single PCIe port.\n"
+						"\tRAID 0 is the only supported configuration for this type of x8 device.\n");
+				break;
+			}
+		}
+	}
 
 	get_dev_size(fd, NULL, &size);
 	/* clear migr_rec when adding disk to container */
-- 
2.7.4

[PATCH] IMSM: Do not update metadata if not able to migrate

[Pawel Baldysiak]

This patch prevents mdadm from updating metadata if migration is
not possible. The same check is done in analyse_change(),
but in that place - metadata is already modified.

Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
---
 super-intel.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/super-intel.c b/super-intel.c
index 6fc9cb1..5c6ab05 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -10067,6 +10067,11 @@ enum imsm_reshape_type imsm_analyze_change(struct supertype *st,
 			pr_err("Error. Chunk size change for RAID 10 is not supported.\n");
 			change = -1;
 			goto analyse_change_exit;
+		} else if (info.component_size % (geo->chunksize/512)) {
+			pr_err("New chunk size (%dK) does not evenly divide device size (%lluk). Aborting...\n",
+			       geo->chunksize/1024, info.component_size/2);
+			change = -1;
+			goto analyse_change_exit;
 		}
 		change = CH_MIGRATION;
 	} else {
-- 
2.7.4

[PATCH] IMSM: Enable spanning between VMD domains

[Pawel Baldysiak]

Each VMD domain adds additional PCI domain. This patch
enables RAID creation with NVMe drives from different
VMD domains.

Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
---
 super-intel.c | 19 +++----------------
 1 file changed, 3 insertions(+), 16 deletions(-)

diff --git a/super-intel.c b/super-intel.c
index f52e3d4..6fc9cb1 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -567,10 +567,6 @@ static int attach_hba_to_super(struct intel_super *super, struct sys_dev *device
 	if (device->type != hba->type)
 		return 2;
 
-	/* Always forbid spanning between VMD domains (seen as different controllers by mdadm) */
-	if (device->type == SYS_DEV_VMD && !path_attached_to_hba(device->path, hba->path))
-		return 2;
-
 	/* Multiple same type HBAs can be used if they share the same OROM */
 	const struct imsm_orom *device_orom = get_orom_by_device_id(device->dev_id);
 
@@ -2023,10 +2019,10 @@ static int detail_platform_imsm(int verbose, int enumerate_only, char *controlle
 
 	for (entry = orom_entries; entry; entry = entry->next) {
 		if (entry->type == SYS_DEV_VMD) {
+			print_imsm_capability(&entry->orom);
 			for (hba = list; hba; hba = hba->next) {
 				if (hba->type == SYS_DEV_VMD) {
 					char buf[PATH_MAX];
-					print_imsm_capability(&entry->orom);
 					printf(" I/O Controller : %s (%s)\n",
 						vmd_domain_to_controller(hba, buf), get_sys_dev_type(hba->type));
 					if (print_vmd_attached_devs(hba)) {
@@ -2034,9 +2030,9 @@ static int detail_platform_imsm(int verbose, int enumerate_only, char *controlle
 							pr_err("failed to get devices attached to VMD domain.\n");
 						result |= 2;
 					}
-					printf("\n");
 				}
 			}
+			printf("\n");
 			continue;
 		}
 
@@ -5999,14 +5995,6 @@ count_volumes(struct intel_hba *hba, int dpa, int verbose)
 		else
 			return 0;
 
-		/* VMD has one orom entry for all domain, but spanning is not allowed.
-		 * VMD arrays should be counted per domain (controller), so skip
-		 * domains that are not the given one.
-		 */
-		if (hba->type == SYS_DEV_VMD &&
-		   (strncmp(device->path, hba->path, strlen(device->path)) != 0))
-			continue;
-
 		devlist = get_devices(hba_path);
 		/* if no intel devices return zero volumes */
 		if (devlist == NULL)
@@ -9268,8 +9256,7 @@ int validate_container_imsm(struct mdinfo *info)
 			return 1;
 		}
 
-		if (orom != orom2 ||
-		    (hba->type == SYS_DEV_VMD && hba != hba2)) {
+		if (orom != orom2) {
 			pr_err("WARNING - IMSM container assembled with disks under different HBAs!\n"
 				"       This operation is not supported and can lead to data loss.\n");
 			return 1;
-- 
2.7.4

[PATCH] Lib.c: Fix geting devname for devices with long path

[Pawel Baldysiak]

In scenario where VMD is enabled, and "x8" type of NVMe drive is
plugged into PCIe switch - the path will be longer than 200 chars
(additional VMD domain + 2 level of PCIe switches).
This patch makes the buffer big enough to handle this kind of
configurations.

Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
---
 lib.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/lib.c b/lib.c
index 057ee52..b640634 100644
--- a/lib.c
+++ b/lib.c
@@ -64,7 +64,7 @@ int get_mdp_major(void)
 char *devid2kname(int devid)
 {
 	char path[30];
-	char link[200];
+	char link[PATH_MAX];
 	static char devnm[32];
 	char *cp;
 	int n;
-- 
2.7.4

Re: [PATCH] IMSM: Add warning message when x8-type device is used

[Jes Sorensen]

Pawel Baldysiak <pawel.baldysiak@intel.com> writes:
> This patch adds the warning message when x8-type device
> is used with IMSM metadata. x8 device is a special
> NVMe drive - two of them on a single PCIe card.
> This card could be a single point of failure for
> RAID levels different than RAID0. x8 devices have
> serial number ending with "-A/-B" or "-1/-2".
>
> Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
> Reviewed-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
> ---
>  super-intel.c | 42 ++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 42 insertions(+)
Hi Pawel,

When posting a multi-patch series, it would be helpful to have it sent
with appropriate headers staging the set is X/Y patches and a cover
letter.

I normally use -n --cover-letter

> diff --git a/super-intel.c b/super-intel.c
> index 5c6ab05..e860e75 100644
> --- a/super-intel.c
> +++ b/super-intel.c
> @@ -5139,9 +5139,51 @@ static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
>  	rv = imsm_read_serial(fd, devname, dd->serial);
>  	if (rv) {
>  		pr_err("failed to retrieve scsi serial, aborting\n");
> +		if (dd->devname)
> +			free(dd->devname);
>  		free(dd);
>  		abort();
>  	}
> +	if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
> +	   (super->hba->type == SYS_DEV_VMD))) {
> +		int i;
> +		char *devpath = diskfd_to_devpath(fd);
> +		char controller_path[PATH_MAX];
> +
> +		if (!devpath) {
> +			pr_err("failed to get devpath, aborting\n");
> +			if (dd->devname)
> +				free(dd->devname);
> +			free(dd);
> +			return 1;
> +		}
> +
> +		snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
> +		free(devpath);
> +
> +		if (devpath_to_vendor(controller_path) == 0x8086) {
> +			/*
> +			 * If Intel's NVMe drive has serial ended with
> +			 * "-A","-B","-1" or "-2" it means that this is "x8"
> +			 * device (double drive on single PCIe card).
> +			 * User should be warned about potential data loss.
> +			 */
> +			for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
> +				/* Skip empty character at the end */
> +				if (dd->serial[i] == 0)
> +					continue;
> +
> +				if (((dd->serial[i] == 'A') || (dd->serial[i] == 'B') ||
> +				   (dd->serial[i] == '1') || (dd->serial[i] == '2')) &&
> +				   (dd->serial[i-1] == '-'))

Minor nit here - please keep this within 80 characters.

Thanks,
Jes

Re: [PATCH] IMSM: Add warning message when x8-type device is used

[Baldysiak, Pawel]

On Fri, 2016-10-21 at 08:47 -0400, Jes Sorensen wrote:
> Pawel Baldysiak <pawel.baldysiak@intel.com> writes:
> > This patch adds the warning message when x8-type device
> > is used with IMSM metadata. x8 device is a special
> > NVMe drive - two of them on a single PCIe card.
> > This card could be a single point of failure for
> > RAID levels different than RAID0. x8 devices have
> > serial number ending with "-A/-B" or "-1/-2".
> >
> > Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
> > Reviewed-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
> > ---
> >  super-intel.c | 42 ++++++++++++++++++++++++++++++++++++++++++
> >  1 file changed, 42 insertions(+)
> Hi Pawel,
> 
> When posting a multi-patch series, it would be helpful to have it sent
> with appropriate headers staging the set is X/Y patches and a cover
> letter.
> 
> I normally use -n --cover-letter
Hi Jes,

Sorry about that. Those patches supposed to be sent separately, but I use
one git send-mail command to sent them all. Now I know that this was not
the most clever thing to do :) Please treat them as separate patches.

> 
> > diff --git a/super-intel.c b/super-intel.c
> > index 5c6ab05..e860e75 100644
> > --- a/super-intel.c
> > +++ b/super-intel.c
> > @@ -5139,9 +5139,51 @@ static int add_to_super_imsm(struct supertype *st, mdu_disk_info_t *dk,
> >  	rv = imsm_read_serial(fd, devname, dd->serial);
> >  	if (rv) {
> >  		pr_err("failed to retrieve scsi serial, aborting\n");
> > +		if (dd->devname)
> > +			free(dd->devname);
> >  		free(dd);
> >  		abort();
> >  	}
> > +	if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
> > +	   (super->hba->type == SYS_DEV_VMD))) {
> > +		int i;
> > +		char *devpath = diskfd_to_devpath(fd);
> > +		char controller_path[PATH_MAX];
> > +
> > +		if (!devpath) {
> > +			pr_err("failed to get devpath, aborting\n");
> > +			if (dd->devname)
> > +				free(dd->devname);
> > +			free(dd);
> > +			return 1;
> > +		}
> > +
> > +		snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
> > +		free(devpath);
> > +
> > +		if (devpath_to_vendor(controller_path) == 0x8086) {
> > +			/*
> > +			 * If Intel's NVMe drive has serial ended with
> > +			 * "-A","-B","-1" or "-2" it means that this is "x8"
> > +			 * device (double drive on single PCIe card).
> > +			 * User should be warned about potential data loss.
> > +			 */
> > +			for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
> > +				/* Skip empty character at the end */
> > +				if (dd->serial[i] == 0)
> > +					continue;
> > +
> > +				if (((dd->serial[i] == 'A') || (dd->serial[i] == 'B') ||
> > +				   (dd->serial[i] == '1') || (dd->serial[i] == '2')) &&
> > +				   (dd->serial[i-1] == '-'))
> 
> Minor nit here - please keep this within 80 characters.

Would you like me to resend corrected version?
I saw that those lines are over 80 chars, but I kept it this way.
I think that its more readable right now, but if you prefer I can break those lines.

Thanks
Pawel
> 
> Thanks,
> Jes

Re: [PATCH] IMSM: Add warning message when x8-type device is used

[Jes Sorensen]

"Baldysiak, Pawel" <pawel.baldysiak@intel.com> writes:
> On Fri, 2016-10-21 at 08:47 -0400, Jes Sorensen wrote:
>> Pawel Baldysiak <pawel.baldysiak@intel.com> writes:
>> > This patch adds the warning message when x8-type device
>> > is used with IMSM metadata. x8 device is a special
>> > NVMe drive - two of them on a single PCIe card.
>> > This card could be a single point of failure for
>> > RAID levels different than RAID0. x8 devices have
>> > serial number ending with "-A/-B" or "-1/-2".
>> >
>> > Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
>> > Reviewed-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
>> > ---
>> >  super-intel.c | 42 ++++++++++++++++++++++++++++++++++++++++++
>> >  1 file changed, 42 insertions(+)
>> Hi Pawel,
>> 
>> When posting a multi-patch series, it would be helpful to have it sent
>> with appropriate headers staging the set is X/Y patches and a cover
>> letter.
>> 
>> I normally use -n --cover-letter
> Hi Jes,
>
> Sorry about that. Those patches supposed to be sent separately, but I use
> one git send-mail command to sent them all. Now I know that this was not
> the most clever thing to do :) Please treat them as separate patches.

Gotcha! No worries, I can do that!

Would you mind fixing up the first one as per the comment below?

Thanks,
Jes

>> 
>> > diff --git a/super-intel.c b/super-intel.c
>> > index 5c6ab05..e860e75 100644
>> > --- a/super-intel.c
>> > +++ b/super-intel.c
>> > @@ -5139,9 +5139,51 @@ static int add_to_super_imsm(struct
>> > supertype *st, mdu_disk_info_t *dk,
>> >  	rv = imsm_read_serial(fd, devname, dd->serial);
>> >  	if (rv) {
>> >  		pr_err("failed to retrieve scsi serial, aborting\n");
>> > +		if (dd->devname)
>> > +			free(dd->devname);
>> >  		free(dd);
>> >  		abort();
>> >  	}
>> > +	if (super->hba && ((super->hba->type == SYS_DEV_NVME) ||
>> > +	   (super->hba->type == SYS_DEV_VMD))) {
>> > +		int i;
>> > +		char *devpath = diskfd_to_devpath(fd);
>> > +		char controller_path[PATH_MAX];
>> > +
>> > +		if (!devpath) {
>> > +			pr_err("failed to get devpath, aborting\n");
>> > +			if (dd->devname)
>> > +				free(dd->devname);
>> > +			free(dd);
>> > +			return 1;
>> > +		}
>> > +
>> > +		snprintf(controller_path, PATH_MAX-1, "%s/device", devpath);
>> > +		free(devpath);
>> > +
>> > +		if (devpath_to_vendor(controller_path) == 0x8086) {
>> > +			/*
>> > +			 * If Intel's NVMe drive has serial ended with
>> > +			 * "-A","-B","-1" or "-2" it means that this is "x8"
>> > +			 * device (double drive on single PCIe card).
>> > +			 * User should be warned about potential data loss.
>> > +			 */
>> > +			for (i = MAX_RAID_SERIAL_LEN-1; i > 0; i--) {
>> > +				/* Skip empty character at the end */
>> > +				if (dd->serial[i] == 0)
>> > +					continue;
>> > +
>> > + if (((dd->serial[i] == 'A') || (dd->serial[i] == 'B') ||
>> > +    (dd->serial[i] == '1') || (dd->serial[i] == '2')) &&
>> > +				   (dd->serial[i-1] == '-'))
>> 
>> Minor nit here - please keep this within 80 characters.
>
> Would you like me to resend corrected version?
> I saw that those lines are over 80 chars, but I kept it this way.
> I think that its more readable right now, but if you prefer I can
> break those lines.
>
> Thanks
> Pawel
>> 
>> Thanks,
>> Jes

Re: [PATCH] IMSM: Enable spanning between VMD domains

[Jes Sorensen]

Pawel Baldysiak <pawel.baldysiak@intel.com> writes:
> Each VMD domain adds additional PCI domain. This patch
> enables RAID creation with NVMe drives from different
> VMD domains.
>
> Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
> ---
>  super-intel.c | 19 +++----------------
>  1 file changed, 3 insertions(+), 16 deletions(-)

Applied!

Thanks,
Jes

Re: [PATCH] Lib.c: Fix geting devname for devices with long path

[Jes Sorensen]

Pawel Baldysiak <pawel.baldysiak@intel.com> writes:
> In scenario where VMD is enabled, and "x8" type of NVMe drive is
> plugged into PCIe switch - the path will be longer than 200 chars
> (additional VMD domain + 2 level of PCIe switches).
> This patch makes the buffer big enough to handle this kind of
> configurations.
>
> Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
> ---
>  lib.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)

Applied!

Thanks,
Jes