Updated the pNFS Developmental Repository
Subsystems (2)
Hello,
I've just updated the kernel and nfs-utils rpms in the pNFS development repository to include:
* The latest nfs-utils release (pnfs-nfs-utils-1-2-4-rc5)
* The latest pNFS kernel bits (pnfs-all-2.6.38-2011-03-25)
These latest bits are *only* in the Fedora 15 repo. The Fedora 13, 14 repos remained the same and the devel (or rawhide) repo was removed.
Instructions on how to install and use these kernels are at:
http://wiki.linux-nfs.org/wiki/index.php/Fedora_pNFS_Client_Setup
If/when bugs are found with these rpms please enter them
in to the NFS Bugzilla Main Page at:
https://bugzilla.linux-nfs.org/index.cgi
Please use the "pNFS" keyword..
Any question or issues please feel free to contact me...
steved.
Steve Dickson wrote:
Hello,
I've just updated the kernel and nfs-utils rpms in the pNFS development repository to include:
* The latest nfs-utils release (pnfs-nfs-utils-1-2-4-rc5)
* The latest pNFS kernel bits (pnfs-all-2.6.38-2011-03-25)
These latest bits are *only* in the Fedora 15 repo. The Fedora 13, 14 repos remained the same and the devel (or rawhide) repo was removed.
Do these builds include the following patch? I sent it to the list on Feb 23 but don't see it in Benny's repo.
Quoted 1 line
From
d7be3ca498142e1f50a7a4120ac890f2b2e4e2a2Mon Sep 17 00:00:00 2001
From: Jim Rees <[email protected]>
Date: Tue, 22 Feb 2011 18:13:20 -0500
Subject: [PATCH] fix stripe device size calculation for layouts not a multiple of stripe unit
If a complex layout has a stripe volume in it, and that volume has a size that is not a multiple of the stripe unit, the total layout size will be wrong because it is based on the truncated stripe volume size. Fix this by using the truncated size for making the mapped device, and the untruncated size in the table for subsequent devices to use.
Also remove the warning for stripe volumes not a multiple of the stripe width, since this is legal and apparently common.
While we're at it, use variable names consistent with the terminology in rfc5661 section 13.2.
--- utils/blkmapd/device-process.c | 10 ++++------ utils/blkmapd/dm-device.c | 21 +++++---------------- 2 files changed, 9 insertions(+), 22 deletions(-)
utils/blkmapd/device-process.c +4 -6
@@ -295,10 +295,10 @@ decode_blk_volume(uint32_t **pp, uint32_t *end, struct bl_volume *vols, int volncase BLOCK_VOLUME_STRIPE :BLK_READBUF (p, end,8 );READ_SECTOR (vol->param.bv_stripe_unit); -off_t chunksize = vol->param.bv_stripe_unit; +off_t stripe_unit = vol->param.bv_stripe_unit;/* Check limitations imposed by device-mapper */ -if ((chunksize & (chunksize -1 )) !=0 - || chunksize < (off_t) (PAGE_SIZE >>9 )) +if ((stripe_unit & (stripe_unit -1 )) !=0 + || stripe_unit < (off_t) (PAGE_SIZE >>9 ))return -EIO ;BLK_READBUF (p, end,4 );READ32 (vol->bv_vol_n);
@@ -306,7 +306,7 @@ decode_blk_volume(uint32_t **pp, uint32_t *end, struct bl_volume *vols, int volnreturn -EIO ; *array_cnt = vol->bv_vol_n;BL_LOG_INFO ("%s: stripe %d nvols=%d unit=%ld\n" , __func__, voln, - vol->bv_vol_n, (long )chunksize); + vol->bv_vol_n, (long )stripe_unit); status =set_vol_array (&p, end, vols, voln);if (status)return status;
@@ -317,9 +317,7 @@ decode_blk_volume(uint32_t **pp, uint32_t *end, struct bl_volume *vols, int volnreturn -EIO ; } } -/* Truncate size to a stripe unit boundary */ vol->bv_size = vol->bv_vols[0 ]->bv_size * vol->bv_vol_n; - vol->bv_size &= ~(chunksize -1 );break ;case BLOCK_VOLUME_CONCAT :BLK_READBUF (p, end,4 );
utils/blkmapd/dm-device.c +5 -16
@@ -369,7 +369,7 @@ static int dm_device_exists(char *dev_name)/* TODO: check the value for DM_DEV_NAME_LEN, DM_TYPE_LEN, DM_PARAMS_LEN */ uint64_t dm_device_create (struct bl_volume *vols,int num_vols) { -uint64_t size, stripe_unit, stripe_size, nstripes, dev =0 ; +uint64_t size, stripe_unit, dev =0 ;unsigned int count = dev_count;int volnum, i, pos;struct bl_volume *node;
@@ -414,21 +414,10 @@ uint64_t dm_device_create(struct bl_volume *vols, int num_vols)if (!table) goto out; table->offset =0 ; - stripe_unit = node->param.bv_stripe_unit <<9 ; - stripe_size = stripe_unit * node->bv_vol_n; - nstripes = node->bv_size * node->bv_vol_n / stripe_size; -/* Make sure total size is a multiple of stripe size */ - size = node->bv_size; -if (size % stripe_size !=0 ) { -/* XXX Should this be an error? */ -BL_LOG_WARNING ( -"%s: %d units of %llu bytes is not a multiple of %lld stripe size\n" , - __func__, node->bv_vol_n, - (long long unsigned ) node->bv_size, - (long long unsigned ) stripe_size); - size = nstripes * stripe_size; - } - table->size = size; +/* Truncate size to a stripe unit boundary */ + stripe_unit = node->param.bv_stripe_unit; + table->size = + node->bv_size - (node->bv_size % stripe_unit);strcpy (table->target_type,"striped" );sprintf (table->params,"%d %llu %n" , node->bv_vol_n, (long long unsigned ) stripe_unit, &pos);
Signature · 1 line
1.7.1
Quoted 9 lines · Jim Rees
Steve Dickson wrote:
Hello,
I've just updated the kernel and nfs-utils rpms in the pNFS development repository to include:
* The latest nfs-utils release (pnfs-nfs-utils-1-2-4-rc5)
* The latest pNFS kernel bits (pnfs-all-2.6.38-2011-03-25)
Quoted 5 lines
These latest bits are *only* in the Fedora 15 repo. The Fedora 13, 14 repos remained the same and the devel (or rawhide) repo was removed.
Do these builds include the following patch? I sent it to the list on Feb 23 but don't see it in Benny's repo.
If they its not in Benny repo, then they are not in this release....
steved.
Quoted 100 lines
From
d7be3ca498142e1f50a7a4120ac890f2b2e4e2a2Mon Sep 17 00:00:00 2001
From: Jim Rees <[email protected]>
Date: Tue, 22 Feb 2011 18:13:20 -0500
Subject: [PATCH] fix stripe device size calculation for layouts not a multiple of stripe unitIf a complex layout has a stripe volume in it, and that volume has a size that is not a multiple of the stripe unit, the total layout size will be wrong because it is based on the truncated stripe volume size. Fix this by using the truncated size for making the mapped device, and the untruncated size in the table for subsequent devices to use.
Also remove the warning for stripe volumes not a multiple of the stripe width, since this is legal and apparently common.
While we're at it, use variable names consistent with the terminology in rfc5661 section 13.2.
Signed-off-by Jim Rees <[email protected]>--- utils/blkmapd/device-process.c | 10 ++++------ utils/blkmapd/dm-device.c | 21 +++++---------------- 2 files changed, 9 insertions(+), 22 deletions(-)utils/blkmapd/device-process.c +4
-6@@ -295,10 +295,10 @@ decode_blk_volume(uint32_t **pp, uint32_t *end, struct bl_volume *vols, int volncase BLOCK_VOLUME_STRIPE :BLK_READBUF (p, end,8 );READ_SECTOR (vol->param.bv_stripe_unit); -off_t chunksize = vol->param.bv_stripe_unit; +off_t stripe_unit = vol->param.bv_stripe_unit;/* Check limitations imposed by device-mapper */ -if ((chunksize & (chunksize -1 )) !=0 - || chunksize < (off_t) (PAGE_SIZE >>9 )) +if ((stripe_unit & (stripe_unit -1 )) !=0 + || stripe_unit < (off_t) (PAGE_SIZE >>9 ))return -EIO ;BLK_READBUF (p, end,4 );READ32 (vol->bv_vol_n);@@ -306,7 +306,7 @@ decode_blk_volume(uint32_t **pp, uint32_t *end, struct bl_volume *vols, int volnreturn -EIO ; *array_cnt = vol->bv_vol_n;BL_LOG_INFO ("%s: stripe %d nvols=%d unit=%ld\n" , __func__, voln, - vol->bv_vol_n, (long )chunksize); + vol->bv_vol_n, (long )stripe_unit); status =set_vol_array (&p, end, vols, voln);if (status)return status;@@ -317,9 +317,7 @@ decode_blk_volume(uint32_t **pp, uint32_t *end, struct bl_volume *vols, int volnreturn -EIO ; } } -/* Truncate size to a stripe unit boundary */ vol->bv_size = vol->bv_vols[0 ]->bv_size * vol->bv_vol_n; - vol->bv_size &= ~(chunksize -1 );break ;case BLOCK_VOLUME_CONCAT :BLK_READBUF (p, end,4 );utils/blkmapd/dm-device.c +5
-16@@ -369,7 +369,7 @@ static int dm_device_exists(char *dev_name)/* TODO: check the value for DM_DEV_NAME_LEN, DM_TYPE_LEN, DM_PARAMS_LEN */ uint64_t dm_device_create (struct bl_volume *vols,int num_vols) { -uint64_t size, stripe_unit, stripe_size, nstripes, dev =0 ; +uint64_t size, stripe_unit, dev =0 ;unsigned int count = dev_count;int volnum, i, pos;struct bl_volume *node;@@ -414,21 +414,10 @@ uint64_t dm_device_create(struct bl_volume *vols, int num_vols)if (!table) goto out; table->offset =0 ; - stripe_unit = node->param.bv_stripe_unit <<9 ; - stripe_size = stripe_unit * node->bv_vol_n; - nstripes = node->bv_size * node->bv_vol_n / stripe_size; -/* Make sure total size is a multiple of stripe size */ - size = node->bv_size; -if (size % stripe_size !=0 ) { -/* XXX Should this be an error? */ -BL_LOG_WARNING ( -"%s: %d units of %llu bytes is not a multiple of %lld stripe size\n" , - __func__, node->bv_vol_n, - (long long unsigned ) node->bv_size, - (long long unsigned ) stripe_size); - size = nstripes * stripe_size; - } - table->size = size; +/* Truncate size to a stripe unit boundary */ + stripe_unit = node->param.bv_stripe_unit; + table->size = + node->bv_size - (node->bv_size % stripe_unit);strcpy (table->target_type,"striped" );sprintf (table->params,"%d %llu %n" , node->bv_vol_n, (long long unsigned ) stripe_unit, &pos);
Quoted 2 lines · Steve Dickson
Do these builds include the following patch? I sent it to the list on Feb 23 but don't see it in Benny's repo.
If they its not in Benny repo, then they are not in this release....
Looks like it is in. My mistake. I think I did a fetch but forgot to reset and/or merge. Sorry for the noise.