megactl/debian/patches/003-Fix_disk_enumeration_loop.dpatch
2017-04-22 15:01:27 +02:00

65 lines
2.8 KiB
Plaintext
Executable File
Raw Permalink Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

#! /bin/sh /usr/share/dpatch/dpatch-run
## 003-Fix_disk_enumeration_loop.dpatch by Pat Suwalski <pat@suwalski.net>
##
## DP: When my controller has a failed disk, megasasctl fails.
## DP:
## DP: I have tracked the problem down to a bad assumption, that the various
## DP: disk enumeration loops would use the number of currently online drives
## DP: as the basis for the iterator.
## DP:
## DP: This number decreases when a disk fails. It is not noticeable in many
## DP: cases, because on many some controllers pinfo->pd_present_count is
## DP: one higher than the actual number of disks.
## DP: However, the same situation would be encountered in a RAID1 system with
## DP: two failed disks.
## DP:
## DP: The attached one-line patch uses the number of ports
## DP: on the controller as the basis.
@DPATCH@
This patch fixes a segfault based on an incorrect assumption that the number
of physical drives to loop over is the number of currently online drives
connected to the controller. This change plays it safe and makes it use the
number of ports on the controller, with a fallback for the worst case
scenario.
Pat Suwalski <pat@suwalski.net>
diff -ur megactl-code/src/adapter.c megactl-code2/src/adapter.c
--- a/src/adapter.c 2014-01-20 11:13:48.114598462 -0500
+++ b/src/adapter.c 2014-01-20 14:33:32.950851825 -0500
@@ -706,7 +706,31 @@
}
qsort (a->channel, a->num_channels, sizeof (*a->channel), cmpChannel);
- a->num_physicals = pinfo->pd_present_count;
+ /* Some notes:
+ Different meanings on different models.
+ - FC_MAX_PHYSICAL_DEVICES used on older controllers, which is 256
+ disks (overallocation)
+ - pd_disk_present_count is number of working drives, not counting
+ missing drives
+ - pd_present_count is unclear. It is pd_disk_present_count + 1 on some
+ controllers
+ - device_interface.port_count contains number of physical ports on the
+ controller
+
+ pd_present_count was used here, but in some controllers causes segfaults
+ when there is a failed drive, and not enough space is allocated.
+
+ Since there cannot be more devices than there are ports, that is a safe
+ number to set without going overboard.
+ */
+ a->num_physicals = pinfo->device_interface.port_count;
+
+ /* On some controllers, namely the PERC6e, the controller does not know
+ how many ports there are in the enclosure. Fall back to the worst case
+ scenario. */
+ if (a->num_physicals < pinfo->pd_disk_present_count)
+ a->num_physicals = FC_MAX_PHYSICAL_DEVICES;
+
if ((a->physical = (struct physical_drive_info *) malloc (a->num_physicals * sizeof (*a->physical))) == NULL)
return "out of memory (physical drives)";
memset (a->physical, 0, a->num_physicals * sizeof (*a->physical));