Hello Karel & list,
Stratus encountered a problem recently when installing RHEL7 onto a
Seagate ST300MP0065 disk (4k native sector size). This disk happened to
contain a GPT from a previous OS install and Anaconda hung when running
"wipefs -f -a /dev/sda" to try and clear the old GPT.
Tests running the same install (old GPT) with 512 sector size disks
succeeded.
To manually reproduce with the latest upstream git util-linux and without
Anaconda we did the following:
% lsscsi | grep /dev/sda
[3:0:1:0] disk SEAGATE ST300MP0065 K001 /dev/sda
% dmesg | grep '3:0:1:0'
scsi 3:0:1:0: Direct-Access SEAGATE ST300MP0065 K001 PQ: 0 ANSI: 6
scsi 3:0:1:0: SSP: handle(0x000a), sas_addr(0x5000c5007687000d), phy(4), device_name(0x5000c5007687000c)
scsi 3:0:1:0: SSP: enclosure_logical_id(0x500605b006d3a670), slot(7)
scsi 3:0:1:0: qdepth(254), tagged(1), simple(0), ordered(0), scsi_level(7), cmd_que(1)
sd 3:0:1:0: [sda] Spinning up disk...
sd 3:0:1:0: Attached scsi generic sg6 type 0
sd 3:0:1:0: [sda] 73259046 4096-byte logical blocks: (300 GB/279 GiB)
sd 3:0:1:0: [sda] Write Protect is off
sd 3:0:1:0: [sda] Mode Sense: db 00 10 08
sd 3:0:1:0: [sda] Write cache: enabled, read cache: enabled, supports DPO and FUA
sd 3:0:1:0: [sda] 73259046 4096-byte logical blocks: (300 GB/279 GiB)
sd 3:0:1:0: [sda] 73259046 4096-byte logical blocks: (300 GB/279 GiB)
sd 3:0:1:0: [sda] Attached SCSI disk
% dd if=/dev/zero of=/dev/sda bs=1M count=100
100+0 records in
100+0 records out
104857600 bytes (105 MB) copied, 0.688838 s, 152 MB/s
% parted /dev/sda --script -- mklabel gpt
% parted /dev/sda --script -- mkpart linux 0 -1
Warning: The resulting partition is not properly aligned for best performance.
% parted /dev/sda --script -- print
Model: SEAGATE ST300MP0065 (scsi)
Disk /dev/sda: 300GB
Sector size (logical/physical): 4096B/4096B
Partition Table: gpt
Disk Flags:
Number Start End Size File system Name Flags
1 24.6kB 300GB 300GB linux
% wipefs -f -a /dev/sda
/dev/sda: 8 bytes were erased at offset 0x00000200 (gpt): 45 46 49 20 50 41 52 54
/dev/sda: 8 bytes were erased at offset 0x00000200 (gpt): 45 46 49 20 50 41 52 54
/dev/sda: 8 bytes were erased at offset 0x00000200 (gpt): 45 46 49 20 50 41 52 54
/dev/sda: 8 bytes were erased at offset 0x00000200 (gpt): 45 46 49 20 50 41 52 54
/dev/sda: 8 bytes were erased at offset 0x00000200 (gpt): 45 46 49 20 50 41 52 54
... endless loop ...
The problem seemed to arise out of libblkid's probe_gpt_pt routine, where
it calls blkid_probe_set_magic with the "lba << 9". That was fine for the
512 disk, but not the 4k disk -- wipefs would later clear the magic data @
byte 512 (not 4096) and then rescan and find the magic again, looping
endlessly.
See the patch below which moved us past this bug for 4k sector
size disks. I *think* it's the right approach, but I wasn't sure if
further refactoring of code south of the blkid_partitions_need_typeonly
check would be necessary. It's not called by wipefs (or blkid for
that matter), so I wasn't sure how to test it.
Regards,
-- Joe
-->8-- -->8-- -->8-- -->8-- -->8-- -->8-- -->8-- -->8--
>From 0216cbd3ae407a1ca2dc50a32542c81181998fb4 Mon Sep 17 00:00:00 2001
From: Joe Lawrence <joe.lawrence@xxxxxxxxxxx>
Date: Wed, 23 Jul 2014 16:39:26 -0400
Subject: [PATCH] libblkid: use sector size for GPT magic offset
probe_gpt_pt should consider the sector size when setting the offset of
the GPT magic, for it may be 512 or 4k.
Signed-off-by: Joe Lawrence <joe.lawrence@xxxxxxxxxxx>
---
libblkid/src/partitions/gpt.c | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/libblkid/src/partitions/gpt.c b/libblkid/src/partitions/gpt.c
index 7fccd51..0c4ce31 100644
--- a/libblkid/src/partitions/gpt.c
+++ b/libblkid/src/partitions/gpt.c
@@ -303,7 +303,7 @@ static int probe_gpt_pt(blkid_probe pr,
blkid_parttable tab = NULL;
blkid_partlist ls;
uint64_t fu, lu;
- uint32_t ssf, i;
+ uint32_t ssz, ssf, i;
efi_guid_t guid;
int ret;
@@ -329,7 +329,9 @@ static int probe_gpt_pt(blkid_probe pr,
blkid_probe_use_wiper(pr, lba * blkid_probe_get_size(pr), 8);
- if (blkid_probe_set_magic(pr, lba << 9,
+ ssz = blkid_probe_get_sectorsize(pr);
+
+ if (blkid_probe_set_magic(pr, ssz * lba,
sizeof(GPT_HEADER_SIGNATURE_STR) - 1,
(unsigned char *) GPT_HEADER_SIGNATURE_STR))
goto err;
@@ -348,13 +350,13 @@ static int probe_gpt_pt(blkid_probe pr,
if (!ls)
goto nothing;
- tab = blkid_partlist_new_parttable(ls, "gpt", lba << 9);
+ tab = blkid_partlist_new_parttable(ls, "gpt", ssz * lba);
if (!tab)
goto err;
blkid_parttable_set_uuid(tab, (const unsigned char *) &guid);
- ssf = blkid_probe_get_sectorsize(pr) / 512;
+ ssf = ssz / 512;
fu = le64_to_cpu(h->first_usable_lba);
lu = le64_to_cpu(h->last_usable_lba);
--
1.7.10.4
--
To unsubscribe from this list: send the line "unsubscribe util-linux" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
