[Cluster-devel] [PATCH v2 0/3] gfs2: Add new resource group header fields

[Cluster-devel] [PATCH v2 0/3] gfs2: Add new resource group header fields

[Andrew Price]

These patches add new fields to struct gfs2_rgrp. The intention is to improve fsck performance and error checking while reducing reliance on the rindex to be read in order to scan through the resource groups in tools such as fsck.gfs2. This is part of a wider goal to obsolete gfs2's metafs, hence the duplication of fields from the rindex. In testing performance, these patches carry very little-to-no overhead.

This posting will be followed by a patch set for gfs2-utils to make use of the new fields, and set and check them enough for testing purposes.

v2: fixed cpu_to_be64/32 typos and made sure rg_skip of the last rgrp is zeroed

Andrew Price (3):
  gfs2: Add a next-resource-group pointer to resource groups
  gfs2: Add rindex fields to rgrp headers
  gfs2: Add a crc field to resource group headers

 fs/gfs2/rgrp.c                   | 16 +++++++++++++++-
 include/uapi/linux/gfs2_ondisk.h | 15 ++++++++++++---
 2 files changed, 27 insertions(+), 4 deletions(-)

-- 
2.13.6

[Cluster-devel] [PATCH v2 1/3] gfs2: Add a next-resource-group pointer to resource groups

[Andrew Price]

Add a new rg_skip field to struct gfs2_rgrp, replacing __pad. The
rg_skip field has the following meaning:

- If rg_skip is zero, it is considered unset and not useful.
- If rg_skip is non-zero, its value will be the number of blocks between
  this rgrp's address and the next rgrp's address. This can be used as a
  hint by fsck.gfs2 when rebuilding a bad rindex, for example.

This will provide less dependency on the rindex in future, and allow
tools such as fsck.gfs2 to iterate the resource groups without keeping
the rindex around.

The field is updated in gfs2_rgrp_out() so that existing file systems
will have it set. This means that any resource groups that aren't ever
written will not be updated. The final rgrp is a special case as there
is no next rgrp, so it will always have a rg_skip of 0 (unless the fs is
extended).

Before this patch, gfs2_rgrp_out() zeroes the __pad field explicitly, so
the rg_skip field can get set back to 0 in cases where nodes with and
without this patch are mixed in a cluster. In some cases, the field may
bounce between being set by one node and then zeroed by another which
may harm performance slightly, e.g. when two nodes create many small
files. In testing this situation is rare but it becomes more likely as
the filesystem fills up and there are fewer resource groups to choose
from. The problem goes away when all nodes are running with this patch.
Dipping into the space currently occupied by the rg_reserved field would
have resulted in the same problem as it is also explicitly zeroed, so
unfortunately there is no other way around it.

Signed-off-by: Andrew Price <anprice@redhat.com>
---
 fs/gfs2/rgrp.c                   | 6 +++++-
 include/uapi/linux/gfs2_ondisk.h | 5 ++++-
 2 files changed, 9 insertions(+), 2 deletions(-)

diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 95b2a57ded33..3c62a5464f60 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1044,12 +1044,16 @@ static void gfs2_rgrp_in(struct gfs2_rgrpd *rgd, const void *buf)
 
 static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf)
 {
+	struct gfs2_rgrpd *next = gfs2_rgrpd_get_next(rgd);
 	struct gfs2_rgrp *str = buf;
 
 	str->rg_flags = cpu_to_be32(rgd->rd_flags & ~GFS2_RDF_MASK);
 	str->rg_free = cpu_to_be32(rgd->rd_free);
 	str->rg_dinodes = cpu_to_be32(rgd->rd_dinodes);
-	str->__pad = cpu_to_be32(0);
+	if (next == NULL)
+		str->rg_skip = 0;
+	else if (next->rd_addr > rgd->rd_addr)
+		str->rg_skip = cpu_to_be32(next->rd_addr - rgd->rd_addr);
 	str->rg_igeneration = cpu_to_be64(rgd->rd_igeneration);
 	memset(&str->rg_reserved, 0, sizeof(str->rg_reserved));
 }
diff --git a/include/uapi/linux/gfs2_ondisk.h b/include/uapi/linux/gfs2_ondisk.h
index 5156bad77b47..da7a30ddef72 100644
--- a/include/uapi/linux/gfs2_ondisk.h
+++ b/include/uapi/linux/gfs2_ondisk.h
@@ -187,7 +187,10 @@ struct gfs2_rgrp {
 	__be32 rg_flags;
 	__be32 rg_free;
 	__be32 rg_dinodes;
-	__be32 __pad;
+	union {
+		__be32 __pad;
+		__be32 rg_skip; /* Distance to the next rgrp in fs blocks */
+	};
 	__be64 rg_igeneration;
 
 	__u8 rg_reserved[80]; /* Several fields from gfs1 now reserved */
-- 
2.13.6

[Cluster-devel] [PATCH v2 2/3] gfs2: Add rindex fields to rgrp headers

[Andrew Price]

Add rg_data0, rg_data and rg_bitbytes to struct gfs2_rgrp. The fields
are identical to their counterparts in struct gfs2_rindex and are
intended to reduce the use of the rindex. For now the fields are only
written back as the in-memory equivalents in struct gfs2_rgrpd are set
using values from the rindex. However, they are needed at this point so
that userspace can make use of them, allowing a migration away from the
rindex over time.

The new fields take up previously reserved space which was explicitly
zeroed on write so, in clusters with mixed kernels, these fields could
get zeroed after being set and this should not be treated as an error.

Signed-off-by: Andrew Price <anprice@redhat.com>
---
 fs/gfs2/rgrp.c                   | 5 +++++
 include/uapi/linux/gfs2_ondisk.h | 7 ++++++-
 2 files changed, 11 insertions(+), 1 deletion(-)

diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index 3c62a5464f60..f7aa9d5eddf1 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -1040,6 +1040,7 @@ static void gfs2_rgrp_in(struct gfs2_rgrpd *rgd, const void *buf)
 	rgd->rd_free = be32_to_cpu(str->rg_free);
 	rgd->rd_dinodes = be32_to_cpu(str->rg_dinodes);
 	rgd->rd_igeneration = be64_to_cpu(str->rg_igeneration);
+	/* rd_data0, rd_data and rd_bitbytes already set from rindex */
 }
 
 static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf)
@@ -1055,6 +1056,10 @@ static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf)
 	else if (next->rd_addr > rgd->rd_addr)
 		str->rg_skip = cpu_to_be32(next->rd_addr - rgd->rd_addr);
 	str->rg_igeneration = cpu_to_be64(rgd->rd_igeneration);
+	str->rg_data0 = cpu_to_be64(rgd->rd_data0);
+	str->rg_data = cpu_to_be32(rgd->rd_data);
+	str->rg_bitbytes = cpu_to_be32(rgd->rd_bitbytes);
+
 	memset(&str->rg_reserved, 0, sizeof(str->rg_reserved));
 }
 
diff --git a/include/uapi/linux/gfs2_ondisk.h b/include/uapi/linux/gfs2_ondisk.h
index da7a30ddef72..648e0cbca574 100644
--- a/include/uapi/linux/gfs2_ondisk.h
+++ b/include/uapi/linux/gfs2_ondisk.h
@@ -192,8 +192,13 @@ struct gfs2_rgrp {
 		__be32 rg_skip; /* Distance to the next rgrp in fs blocks */
 	};
 	__be64 rg_igeneration;
+	/* The following 3 fields are duplicated from gfs2_rindex to reduce
+	   reliance on the rindex */
+	__be64 rg_data0;     /* First data location */
+	__be32 rg_data;      /* Number of data blocks in rgrp */
+	__be32 rg_bitbytes;  /* Number of bytes in data bitmaps */
 
-	__u8 rg_reserved[80]; /* Several fields from gfs1 now reserved */
+	__u8 rg_reserved[64]; /* Several fields from gfs1 now reserved */
 };
 
 /*
-- 
2.13.6

[Cluster-devel] [PATCH v2 3/3] gfs2: Add a crc field to resource group headers

[Andrew Price]

Add the rg_crc field to store a crc32 of the gfs2_rgrp structure. This
allows us to check resource group headers' integrity and removes the
requirement to check them against the rindex entries in fsck. If this
field is found to be zero, it should be ignored (or updated with an
accurate value).

Signed-off-by: Andrew Price <anprice@redhat.com>
---
 fs/gfs2/rgrp.c                   | 5 +++++
 include/uapi/linux/gfs2_ondisk.h | 3 ++-
 2 files changed, 7 insertions(+), 1 deletion(-)

diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c
index f7aa9d5eddf1..2e471fd8f415 100644
--- a/fs/gfs2/rgrp.c
+++ b/fs/gfs2/rgrp.c
@@ -34,6 +34,7 @@
 #include "log.h"
 #include "inode.h"
 #include "trace_gfs2.h"
+#include "dir.h"
 
 #define BFITNOENT ((u32)~0)
 #define NO_BLOCK ((u64)~0)
@@ -1047,6 +1048,7 @@ static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf)
 {
 	struct gfs2_rgrpd *next = gfs2_rgrpd_get_next(rgd);
 	struct gfs2_rgrp *str = buf;
+	u32 crc;
 
 	str->rg_flags = cpu_to_be32(rgd->rd_flags & ~GFS2_RDF_MASK);
 	str->rg_free = cpu_to_be32(rgd->rd_free);
@@ -1059,6 +1061,9 @@ static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf)
 	str->rg_data0 = cpu_to_be64(rgd->rd_data0);
 	str->rg_data = cpu_to_be32(rgd->rd_data);
 	str->rg_bitbytes = cpu_to_be32(rgd->rd_bitbytes);
+	str->rg_crc = 0;
+	crc = gfs2_disk_hash(buf, sizeof(struct gfs2_rgrp));
+	str->rg_crc = cpu_to_be32(crc);
 
 	memset(&str->rg_reserved, 0, sizeof(str->rg_reserved));
 }
diff --git a/include/uapi/linux/gfs2_ondisk.h b/include/uapi/linux/gfs2_ondisk.h
index 648e0cbca574..09f0920f07e9 100644
--- a/include/uapi/linux/gfs2_ondisk.h
+++ b/include/uapi/linux/gfs2_ondisk.h
@@ -197,8 +197,9 @@ struct gfs2_rgrp {
 	__be64 rg_data0;     /* First data location */
 	__be32 rg_data;      /* Number of data blocks in rgrp */
 	__be32 rg_bitbytes;  /* Number of bytes in data bitmaps */
+	__be32 rg_crc;       /* crc32 of the structure with this field 0 */
 
-	__u8 rg_reserved[64]; /* Several fields from gfs1 now reserved */
+	__u8 rg_reserved[60]; /* Several fields from gfs1 now reserved */
 };
 
 /*
-- 
2.13.6

[Cluster-devel] [PATCH v2 0/3] gfs2: Add new resource group header fields

[Bob Peterson]

----- Original Message -----
| These patches add new fields to struct gfs2_rgrp. The intention is to improve
| fsck performance and error checking while reducing reliance on the rindex to
| be read in order to scan through the resource groups in tools such as
| fsck.gfs2. This is part of a wider goal to obsolete gfs2's metafs, hence the
| duplication of fields from the rindex. In testing performance, these patches
| carry very little-to-no overhead.
| 
| This posting will be followed by a patch set for gfs2-utils to make use of
| the new fields, and set and check them enough for testing purposes.
| 
| v2: fixed cpu_to_be64/32 typos and made sure rg_skip of the last rgrp is
| zeroed
| 
| Andrew Price (3):
|   gfs2: Add a next-resource-group pointer to resource groups
|   gfs2: Add rindex fields to rgrp headers
|   gfs2: Add a crc field to resource group headers
| 
|  fs/gfs2/rgrp.c                   | 16 +++++++++++++++-
|  include/uapi/linux/gfs2_ondisk.h | 15 ++++++++++++---
|  2 files changed, 27 insertions(+), 4 deletions(-)
| 
| --
| 2.13.6

Hi,

Series ACK

Bob Peterson
Red Hat File Systems

[Cluster-devel] [PATCH v2 0/3] gfs2: Add new resource group header fields

[Bob Peterson]

----- Original Message -----
| These patches add new fields to struct gfs2_rgrp. The intention is to improve
| fsck performance and error checking while reducing reliance on the rindex to
| be read in order to scan through the resource groups in tools such as
| fsck.gfs2. This is part of a wider goal to obsolete gfs2's metafs, hence the
| duplication of fields from the rindex. In testing performance, these patches
| carry very little-to-no overhead.
| 
| This posting will be followed by a patch set for gfs2-utils to make use of
| the new fields, and set and check them enough for testing purposes.
| 
| v2: fixed cpu_to_be64/32 typos and made sure rg_skip of the last rgrp is
| zeroed
| 
| Andrew Price (3):
|   gfs2: Add a next-resource-group pointer to resource groups
|   gfs2: Add rindex fields to rgrp headers
|   gfs2: Add a crc field to resource group headers
| 
|  fs/gfs2/rgrp.c                   | 16 +++++++++++++++-
|  include/uapi/linux/gfs2_ondisk.h | 15 ++++++++++++---
|  2 files changed, 27 insertions(+), 4 deletions(-)
| 
| --
| 2.13.6

Hi,

Thanks. All three patches are now pushed to the for-next branch of the linux-gfs2 tree:

gfs2: Add a next-resource-group pointer to resource groups
https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=65adc27375a85beb57c3869cedb2a410fad9c288

gfs2: Add rindex fields to rgrp headers
https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=166725d96322473305e35f9d580591a01697ab29

gfs2: Add a crc field to resource group headersfor-next
https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=850d2d915fa69011bef9bd668499cce889fdd8b3

Regards,

Bob Peterson
Red Hat File Systems

[Cluster-devel] [PATCH v2 0/3] gfs2: Add new resource group header fields

[Andrew Price]

Hi Bob,

On 08/12/17 15:14, Bob Peterson wrote:
> ----- Original Message -----
> | These patches add new fields to struct gfs2_rgrp. The intention is to improve
> | fsck performance and error checking while reducing reliance on the rindex to
> | be read in order to scan through the resource groups in tools such as
> | fsck.gfs2. This is part of a wider goal to obsolete gfs2's metafs, hence the
> | duplication of fields from the rindex. In testing performance, these patches
> | carry very little-to-no overhead.
> |
> | This posting will be followed by a patch set for gfs2-utils to make use of
> | the new fields, and set and check them enough for testing purposes.
> |
> | v2: fixed cpu_to_be64/32 typos and made sure rg_skip of the last rgrp is
> | zeroed
> |
> | Andrew Price (3):
> |   gfs2: Add a next-resource-group pointer to resource groups
> |   gfs2: Add rindex fields to rgrp headers
> |   gfs2: Add a crc field to resource group headers
> |
> |  fs/gfs2/rgrp.c                   | 16 +++++++++++++++-
> |  include/uapi/linux/gfs2_ondisk.h | 15 ++++++++++++---
> |  2 files changed, 27 insertions(+), 4 deletions(-)
> |
> | --
> | 2.13.6
> 
> Hi,
> 
> Series ACK
> 
> Bob Peterson
> Red Hat File Systems

I've just realised that these aren't in for-next despite the positive 
response. Can they get in for the upcoming merge window?

Andy

[Cluster-devel] [PATCH v2 0/3] gfs2: Add new resource group header fields

[Bob Peterson]

Hi Andy,

----- Original Message -----
| I've just realised that these aren't in for-next despite the positive
| response. Can they get in for the upcoming merge window?

Your three patches are indeed already included in for-next and
have been for some time. I'm not sure why you're not seeing them,
but I needed to do several push --force operations to fix various
patch reworks, screw-ups, and dependencies. Perhaps you just need
to git pull again?

Below are links to the patches in for-next:

| Andrew Price (3):
|   gfs2: Add a next-resource-group pointer to resource groups
https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=65adc27375a85beb57c3869cedb2a410fad9c288

|   gfs2: Add rindex fields to rgrp headers
https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=166725d96322473305e35f9d580591a01697ab29

|   gfs2: Add a crc field to resource group headers
https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=850d2d915fa69011bef9bd668499cce889fdd8b3

So yes, they should go into the next merge window.

Regards,

Bob Peterson
Red Hat File Systems

[Cluster-devel] [PATCH v2 0/3] gfs2: Add new resource group header fields

[Andrew Price]

On 29/01/18 13:30, Bob Peterson wrote:
> Hi Andy,
> 
> ----- Original Message -----
> | I've just realised that these aren't in for-next despite the positive
> | response. Can they get in for the upcoming merge window?
> 
> Your three patches are indeed already included in for-next and
> have been for some time. I'm not sure why you're not seeing them,
> but I needed to do several push --force operations to fix various
> patch reworks, screw-ups, and dependencies. Perhaps you just need
> to git pull again?
> 
> Below are links to the patches in for-next:
> 
> | Andrew Price (3):
> |   gfs2: Add a next-resource-group pointer to resource groups
> https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=65adc27375a85beb57c3869cedb2a410fad9c288
> 
> |   gfs2: Add rindex fields to rgrp headers
> https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=166725d96322473305e35f9d580591a01697ab29
> 
> |   gfs2: Add a crc field to resource group headers
> https://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2.git/commit/fs/gfs2?h=for-next&id=850d2d915fa69011bef9bd668499cce889fdd8b3
> 
> So yes, they should go into the next merge window.

Ah I see them in the log now, thanks. They must have been pushed further 
down the log than the bulk of the other patches due to the punch-hole 
merge and I missed them. Sorry for the noise.

Andy