| Message ID | 20090722080129.E9393D340B@pentland.suse.de (mailing list archive) |
|---|---|
| State | Deferred, archived |
| Headers | show |
On Wed, 2009-07-22 at 10:01 +0200, Hannes Reinecke wrote: > Worse, it might trigger I/O errors as the partition table might not > be accessible (eg for the passive path of a multipath device) or > even invalid (eg for RAID0 dmraid). Not that it removes the need or motivation for this patch but the case of an array partition table being mis-interpreted by the kernel for a RAID member device should no longer trigger I/O errors. Kernels since 2.6.27 should truncate any over-size partitions to the size of the device: commit ac0d86f5809598ddcd6bfa0ea8245ccc910e9eac Author: Kay Sievers <kay.sievers@vrfy.org> Date: Wed Oct 15 22:04:21 2008 -0700 block: sanitize invalid partition table entries We currently follow blindly what the partition table lies about the disk, and let the kernel create block devices which can not be accessed. Trying to identify the device leads to kernel logs full of: sdb: rw=0, want=73392, limit=28800 attempt to access beyond end of device [snip] I've tested this patch with a few dmraid setups that previously spat I/O errors whenever something probed a member device. Although this isn't exactly a solution it does quiet down the log noise (I'd thought about submitting a patch to dmraid to issue a BLKPG_DEL_PARTITION for each partition on each member device that it discovers but didn't get around to doing anything about it yet). Regards, Bryn. -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel
On Wed, 2009-07-22 at 17:53 +0100, Bryn M. Reeves wrote: > On Wed, 2009-07-22 at 10:01 +0200, Hannes Reinecke wrote: > > Worse, it might trigger I/O errors as the partition table might not > > be accessible (eg for the passive path of a multipath device) or > > even invalid (eg for RAID0 dmraid). > > Not that it removes the need or motivation for this patch but the case > of an array partition table being mis-interpreted by the kernel for a > RAID member device should no longer trigger I/O errors. Kernels since > 2.6.27 should truncate any over-size partitions to the size of the > device: > > commit ac0d86f5809598ddcd6bfa0ea8245ccc910e9eac > Author: Kay Sievers <kay.sievers@vrfy.org> > Date: Wed Oct 15 22:04:21 2008 -0700 > > block: sanitize invalid partition table entries > > We currently follow blindly what the partition table lies about the > disk, and let the kernel create block devices which can not be > accessed. > Trying to identify the device leads to kernel logs full of: > sdb: rw=0, want=73392, limit=28800 > attempt to access beyond end of device > [snip] > > I've tested this patch with a few dmraid setups that previously spat I/O > errors whenever something probed a member device. Although this isn't > exactly a solution it does quiet down the log noise (I'd thought about > submitting a patch to dmraid to issue a BLKPG_DEL_PARTITION for each > partition on each member device that it discovers but didn't get around > to doing anything about it yet). We've got that in RHEL/Fedora dmraid already with option "--rm_partitions". Regards, Heinz > > Regards, > Bryn. > > > -- > dm-devel mailing list > dm-devel@redhat.com > https://www.redhat.com/mailman/listinfo/dm-devel -- dm-devel mailing list dm-devel@redhat.com https://www.redhat.com/mailman/listinfo/dm-devel
diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt index d08759a..47bed29 100644 --- a/Documentation/kernel-parameters.txt +++ b/Documentation/kernel-parameters.txt @@ -1424,6 +1424,11 @@ and is between 256 and 4096 characters. It is defined in the file mtdparts= [MTD] See drivers/mtd/cmdlinepart.c. + no_partition_scan [KNL] Inhibit in-kernel partition scan + for all block devices. Partition scan can be + re-enabled for individual devices by writing + a non-zero value in /sys/block/XX/range. + onenand.bdry= [HW,MTD] Flex-OneNAND Boundary Configuration Format: [die0_boundary][,die0_lock][,die1_boundary][,die1_lock] diff --git a/block/genhd.c b/block/genhd.c index f4c64c2..d120ec8 100644 --- a/block/genhd.c +++ b/block/genhd.c @@ -504,6 +504,18 @@ static int exact_lock(dev_t devt, void *data) return 0; } +static int __read_mostly no_partition_scan; + +static int __init no_partition_scan_setup(char *str) +{ + no_partition_scan = 1; + printk(KERN_INFO "genhd: omit partition scan.\n"); + + return 1; +} + +__setup("no_partition_scan", no_partition_scan_setup); + /** * add_disk - add partitioning information to kernel list * @disk: per-device partitioning information @@ -527,6 +539,8 @@ void add_disk(struct gendisk *disk) WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT)); disk->flags |= GENHD_FL_UP; + if (no_partition_scan) + disk->flags |= GENHD_FL_USERSPACE_PARTITIONS; retval = blk_alloc_devt(&disk->part0, &devt); if (retval) { @@ -816,7 +830,27 @@ static ssize_t disk_range_show(struct device *dev, { struct gendisk *disk = dev_to_disk(dev); - return sprintf(buf, "%d\n", disk->minors); + return sprintf(buf, "%d\n", + (disk_userspace_partitions(disk) ? 0 : disk->minors)); +} + +static ssize_t disk_range_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct gendisk *disk = dev_to_disk(dev); + int i; + + if (count > 0 && sscanf(buf, "%d", &i) > 0) { + if (i == 0) + disk->flags |= GENHD_FL_USERSPACE_PARTITIONS; + else if (i <= disk->minors) + disk->flags &= ~GENHD_FL_USERSPACE_PARTITIONS; + else + count = -EINVAL; + } + + return count; } static ssize_t disk_ext_range_show(struct device *dev, @@ -824,7 +858,8 @@ static ssize_t disk_ext_range_show(struct device *dev, { struct gendisk *disk = dev_to_disk(dev); - return sprintf(buf, "%d\n", disk_max_parts(disk)); + return sprintf(buf, "%d\n", + disk_userspace_partitions(disk) ? 0: disk_max_parts(disk)); } static ssize_t disk_removable_show(struct device *dev, @@ -861,7 +896,7 @@ static ssize_t disk_alignment_offset_show(struct device *dev, return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); } -static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL); +static DEVICE_ATTR(range, S_IRUGO, disk_range_show, disk_range_store); static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL); static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL); static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
For some setups (multipath or dmraid) the in-kernel partition scan is pointless as the (block) partitions won't be used anywhere. Worse, it might trigger I/O errors as the partition table might not be accessible (eg for the passive path of a multipath device) or even invalid (eg for RAID0 dmraid). This patch allows to switch off the in-kernel partition scan by adding 'no_partition_scan' to the kernel commandline. Partitions scan can be allowed for individual disk by echoing a positive number into /sys/block/XXX/range and rescan the disk. Signed-off-by: Hannes Reinecke <hare@suse.de> --- Documentation/kernel-parameters.txt | 5 ++++ block/genhd.c | 41 ++++++++++++++++++++++++++++++++-- 2 files changed, 43 insertions(+), 3 deletions(-)