Cmnt: [SRU][PULL][jammy:linux-gcp] Patches needed for AmpereOne (arm64)

Thu Nov 30 21:56:04 UTC 2023

On 11/30/23 2:47 PM, Dimitri John Ledkov wrote:
> On Thu, 30 Nov 2023 at 14:33, John Cabaj <john.cabaj at canonical.com> wrote:
>>
>> BugLink: https://bugs.launchpad.net/bugs/2044192
>>
>> [Impact]
>>
>> * Google requested patches for AmpereOne machine type
> 
> 
> As all of these are from v6.1 kernel and thus well tested by all our
> kernels, and it is even present in our hwe jammy kernels, may I ask
> you to also submit these patches for jammy generic under HWE
> exception?
> 
> Ampere based servers are very popular, and actually most public clouds
> use them for their arm64 instances. Thus it is likely all of them will
> want these patches.
> 
> But for generic, Canonical itself needs or will likely to need them
> for our internal deployments too.
> 
> Submitting this for jammy generic will also ensure that we will be
> able to continue testing generic flavour on gcp arm64 instances (which
> we often do when hunting down regressions and bugs).
> 
> Would it be possible to get these into generic in time for the Jammy
> point release

Good point - I can work on getting them submitted for Jammy generic as well.

> 
>>
>> [Fix]
>>
>> * Cherry-picks from upstream
>>
>> 484baebd076c ("clocksource/drivers/arch_arm_timer: Move workaround synchronisation around")
>> ad6c750791e3 ("clocksource/drivers/arm_arch_timer: Fix masking for high freq counters")
>> 6e69aa090138 ("clocksource/drivers/arm_arch_timer: Drop unnecessary ISB on CVAL programming")
>> ab9a0239324a ("clocksource/drivers/arm_arch_timer: Remove any trace of the TVAL programming interface")
>> 9f074b90e6fd ("clocksource/drivers/arm_arch_timer: Work around broken CVAL implementations")
>> 562eb64c090f ("clocksource/drivers/arm_arch_timer: Advertise 56bit timer to the core code")
>> 7bc14651e24e ("clocksource/drivers/arm_arch_timer: Move MMIO timer programming over to CVAL")
>> 8b06e41fe7bc ("clocksource/drivers/arm_arch_timer: Fix MMIO base address vs callback ordering issue")
>> 0ffa6302e2f6 ("clocksource/drivers/arm_arch_timer: Move drop _tval from erratum function names")
>> 8065137de0cf ("clocksource/drivers/arm_arch_timer: Move system register timer programming over to CVAL")
>> 674941dc157e ("clocksource/drivers/arm_arch_timer: Extend write side of timer register accessors to u64")
>> 0c1f5fc141ea ("clocksource/drivers/arm_arch_timer: Drop CNT*_TVAL read accessors")
>> 02aecc78b278 ("clocksource/arm_arch_timer: Add build-time guards for unhandled register accesses")
>>
>> [Test Case]
>>
>> * Compile tested
>> * Boot tested on AmpereOne instance
>>
>> [Where things could go wrong]
>>
>> * Low chance of regression. Changes isolated to ARM timers.
>>
>> [Other Info]
>>
>> * SF #00372821
>>
>> -----
>>
>> The following changes since commit 04c19cd6345a4c14df5e64d6a4523d52342321dc:
>>
>>   UBUNTU: Ubuntu-gcp-5.15.0-1048.56 (2023-11-18 14:50:34 -0300)
>>
>> are available in the git repository at:
>>
>>   git+ssh://user@kathleen/home/john-cabaj/for-review/jammy_linux-gcp/ ampere_one
> 
> Please do not leak internal private hostnames publically.
> 
> Also this is a public mailing list mailing submission, thus it should
> use publically accessible repositories.
> 
> To make things easy in launchpad, please consider doing a "fork" of
> the linux-gcp repo into your own namespace, and then pushing to it.
> 
> At https://code.launchpad.net/~canonical-kernel/ubuntu/+source/linux-gcp/+git/jammy/+fork
> you can fork linux-gcp jammy repo which will create a similar looking
> URL but under ~your-lp-account name, which you will then be able to
> efficiently push to.
> 
> And for all private kernels, we should scope them under private
> projects such that personal forks work there too, without any
> information leaks.

Noted.

> 
>>
>> for you to fetch changes up to 484baebd076c93b60979ae26718477c8bf1bb47a:
>>
>>   clocksource/drivers/arch_arm_timer: Move workaround synchronisation around (2023-11-21 16:21:06 -0600)
>>
>> ----------------------------------------------------------------
>> Marc Zyngier (12):
>>       clocksource/arm_arch_timer: Add build-time guards for unhandled register accesses
>>       clocksource/drivers/arm_arch_timer: Drop CNT*_TVAL read accessors
>>       clocksource/drivers/arm_arch_timer: Extend write side of timer register accessors to u64
>>       clocksource/drivers/arm_arch_timer: Move system register timer programming over to CVAL
>>       clocksource/drivers/arm_arch_timer: Move drop _tval from erratum function names
>>       clocksource/drivers/arm_arch_timer: Fix MMIO base address vs callback ordering issue
>>       clocksource/drivers/arm_arch_timer: Move MMIO timer programming over to CVAL
>>       clocksource/drivers/arm_arch_timer: Advertise 56bit timer to the core code
>>       clocksource/drivers/arm_arch_timer: Work around broken CVAL implementations
>>       clocksource/drivers/arm_arch_timer: Remove any trace of the TVAL programming interface
>>       clocksource/drivers/arm_arch_timer: Drop unnecessary ISB on CVAL programming
>>       clocksource/drivers/arch_arm_timer: Move workaround synchronisation around
>>
>> Oliver Upton (1):
>>       clocksource/drivers/arm_arch_timer: Fix masking for high freq counters
>>
>>  arch/arm/include/asm/arch_timer.h    |  37 +++++++++++++---------
>>  arch/arm64/include/asm/arch_timer.h  |  52 ++++++++++++++----------------
>>  drivers/clocksource/arm_arch_timer.c | 243 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-------------------------------------------------------
>>  include/clocksource/arm_arch_timer.h |   2 +-
>>  4 files changed, 193 insertions(+), 141 deletions(-)
>>
>> --
>> kernel-team mailing list
>> kernel-team at lists.ubuntu.com
>> https://lists.ubuntu.com/mailman/listinfo/kernel-team
> 
> 
> 

John