lshw does not display cpu information correctly on aarch64
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
lshw (Ubuntu) |
Fix Released
|
High
|
Manoj Iyer | ||
Trusty |
Fix Released
|
Undecided
|
Unassigned | ||
Vivid |
Fix Released
|
Undecided
|
Unassigned | ||
Wily |
Fix Released
|
High
|
Manoj Iyer |
Bug Description
On certain aarch64 systems lshw does not display cpu information correctly. It displays cpu0 and the any subsequent cpus are displayed as DISABLED. On a multi core aarch64 system lshw displays the following:
*-core
description: Motherboard
physical id: 0
*-cpu:0
product: cpu
physical id: 0
bus info: cpu@0
*-cpu:1 DISABLED
product: cpu
physical id: 2
bus info: cpu@1
*-cpu:2 DISABLED
product: cpu
physical id: 4
bus info: cpu@2
*-cpu:3 DISABLED
product: cpu
physical id: 5
bus info: cpu@3
Please consider the attached merge proposal for SRU in trusty.
[Impact]
* When users run lshw on aarch64 systems the CPU information is not displayed correctly. It displays cpu0 and the any subsequent cpus are displayed as DISABLED. Please refer to bug description for details on output.
* This bug can be fixed by adding code to handle the aarch64 case similar to the existing code for powerpc, x86, x64, ia64 etc. But, lshw upstream has done the work of adding support for s390 and aarch64. It will be better to backport those changes rather than using a bandaid fix for aarch64 in ubuntu.
[Test Case]
* On an ubuntu aarch64 system run the command sudo lshw and look at the information displayed for cpu. Without the patch you should see that CPU 1-N will be DISABLED.
* With the patch run lshw on aarch64 ubuntu system and you should see that it lists CPUs 1-N and display information about the CPU.
[Regression Potential]
* Currently support for aarch64 is broken, so if a regression were to occur it could be in other archs. But, the changes introduced by this patch is to support aarch64 and s390 systems and those changes should not impact other archs.
* This patch was tested on a 48 core aarch64 system running Trusty and found to work as expected. lshw package for wily and trusty were built and tested on an aarch64 system running Trusty.
Related branches
- Michael Terry: Approve
-
Diff: 253 lines (+238/-0)2 files modifieddebian/patches/cpuinfo-arm-aarch64-s390-support.patch (+237/-0)
debian/patches/series (+1/-0)
- Serge Hallyn: Approve
-
Diff: 255 lines (+240/-0)2 files modifieddebian/patches/cpuinfo-arm-aarch64-s390-support.patch (+239/-0)
debian/patches/series (+1/-0)
- dann frazier: Needs Fixing
-
Diff: 115 lines (+95/-0)3 files modifieddebian/changelog (+6/-0)
debian/patches/cpuinfo-arm-aarch64-support.patch (+88/-0)
debian/patches/series (+1/-0)
Changed in lshw (Ubuntu): | |
assignee: | nobody → Manoj Iyer (manjo) |
importance: | Undecided → High |
Changed in lshw (Ubuntu): | |
status: | New → In Progress |
milestone: | none → ubuntu-15.10 |
description: | updated |
description: | updated |
Changed in lshw (Ubuntu Wily): | |
status: | In Progress → Fix Released |
Changed in lshw (Ubuntu Vivid): | |
status: | New → Confirmed |
Changed in lshw (Ubuntu Trusty): | |
status: | New → Confirmed |
Changed in lshw (Ubuntu Vivid): | |
status: | Confirmed → In Progress |
@Manoj: can you MP a fix for vivid before we do the trusty one? Basically like the wily one, but with the following changes:
- Please drop the changes to s390x - it won't be supported in trusty, an SRUs should avoid unnecessary changes. (value) ) patches/ series file
- Please either drop the changes on arm32 *or* verify the bug on an arm32 platform and verify the fix there.
- Please drop the arch-independent changes (i.e. cpu->setProduct
- Please drop the __ID("@(#) change at the top of the file - it is an unnecessary change.
- Please make the new patch the last entry in the debian/
- Please add a debian/changelog entry - ideally copying the one from wily, but with an LP bug reference.
Once we SRU the vivid one, we can move on to trusty with the same above changes.