LinuxPPC-Dev Archive on lore.kernel.org
 help / color / Atom feed
From: Hari Bathini <hbathini@linux.ibm.com>
To: Michael Ellerman <mpe@ellerman.id.au>,
	"Oliver O'Halloran" <oohall@gmail.com>
Cc: Ananth N Mavinakayanahalli <ananth@linux.ibm.com>,
	Mahesh J Salgaonkar <mahesh@linux.ibm.com>,
	Nicholas Piggin <npiggin@gmail.com>,
	linuxppc-dev <linuxppc-dev@ozlabs.org>,
	Vasant Hegde <hegdevasant@linux.ibm.com>,
	Daniel Axtens <dja@axtens.net>
Subject: Re: [PATCH v5 21/31] powernv/fadump: process architected register state data provided by firmware
Date: Tue, 10 Sep 2019 21:40:44 +0530
Message-ID: <8b26238d-d8f6-c9a5-714e-b4682b94f8a5@linux.ibm.com> (raw)
In-Reply-To: <87sgp4z1ov.fsf@mpe.ellerman.id.au>



On 10/09/19 7:35 PM, Michael Ellerman wrote:
> Hari Bathini <hbathini@linux.ibm.com> writes:
>> On 09/09/19 9:03 PM, Oliver O'Halloran wrote:
>>> On Mon, Sep 9, 2019 at 11:23 PM Hari Bathini <hbathini@linux.ibm.com> wrote:
>>>> On 04/09/19 5:50 PM, Michael Ellerman wrote:
>>>>> Hari Bathini <hbathini@linux.ibm.com> writes:
>>>> [...]
>>>>
>>>>>> +/*
>>>>>> + * CPU state data is provided by f/w. Below are the definitions
>>>>>> + * provided in HDAT spec. Refer to latest HDAT specification for
>>>>>> + * any update to this format.
>>>>>> + */
>>>>>
>>>>> How is this meant to work? If HDAT ever changes the format they will
>>>>> break all existing kernels in the field.
>>>>>
>>>>>> +#define HDAT_FADUMP_CPU_DATA_VERSION                1
>>>>
>>>> Changes are not expected here. But this is just to cover for such scenario,
>>>> if that ever happens.
>>>
>>> The HDAT spec doesn't define the SPR numbers for NIA, MSR and the CR.
>>> As far as I can tell the values you've assumed here are chip-specific,
>>> non-architected SPR numbers that come from an array buried somewhere
>>> in the SBE codebase. I don't believe you for a second when you say
>>> that this will never change.
>>
>> At least, the understanding is that this numbers not change across processor
>> generations. If something changes, it is supposed to be handled in SBE. Also,
>> I am told this numbers would be listed in the HDAT Spec. Not sure if that
>> happened yet though. Vasant, you have anything to add?
> 
> That doesn't help much because the HDAT spec is not public.
> 
> The point is with the code written the way it is, these values *must
> not* change, or else all existing kernels will be broken, which is not
> acceptable.

Yeah. It is absurd to error out just by looking at version number...

> 
>>>> Also, I think it is a bit far-fetched to error out if versions mismatch.
>>>> Warning and proceeding sounds worthier because the changes are usually
>>>> backward compatible, if and when there are any. Will update accordingly...
>>>
>>> Literally the only reason I didn't drop the CPU DATA parts of the OPAL
>>> MPIPL series was because I assumed the kernel would do the sensible
>>> thing and reject or ignore the structure if it did not know how to
>>> parse the data.
>>
>> I think, the changes if any, would have to be backward compatible for the sake
>> of sanity.
> 
> People need to understand that this is an ABI between firmware and
> in-the-field distribution kernels which are only updated at customer
> discretion, or possibly never.
> 
> Any changes *must be* backward compatible.
> 
> Looking at the header struct:
> 
> +struct hdat_fadump_thread_hdr {
> +	__be32  pir;
> +	/* 0x00 - 0x0F - The corresponding stop state of the core */
> +	u8      core_state;
> +	u8      reserved[3];
> 
> You have those 3 reserved bytes, so a future revision could repurpose
> one of those as a flag to indicate a new format. And/or the hdr could be
> made bigger and new kernels could be taught to look for new things in
> the space after the hdr but before the reg entries.
> 
> So I think there is a reasonable mechanism for extending the format in
> future, but my point is people must understand that this is an ABI and
> changes must be made accordingly.

True. The folks who make the changes to this format should be aware that
breaking kernel ABI is not going to be pretty and I think they are :)

> 
>> Even if they are not, we are better off exporting the /proc/vmcore
>> with a warning and some crazy CPU register data (if parsing goes alright) than
>> no dump at all? 
> 
> If it's just a case of reg entries that we don't recognise then yes I
> think it would be OK to just skip them and continue exporting. But if
> there's any deeper misunderstanding of the format then we should bail
> out.

Sure. Will try and fix that by first trying to do a sanity check on the
fields that are needed for parsing the data and proceed with a warning if
nothing weird is detected and fallback to just appending crashing cpu as
done in patch 16/31, if anything weird is observed. That should hopefully
take care of all cases in the best possible way..

> 
> I notice now that you don't do anything in opal_fadump_set_regval_regnum()
> if you are passed a register we don't understand, so that probably needs
> fixing.

f/w provides about 100 odd registers in the CPU state data. Most of them
pt_regs doesn't care about. So, opal_fadump_set_regval_regnum is happy as
long as it find the registers listed in it. Unless, pt_regs changes, we
can stick to this and ignore the rest of them?

- Hari


  reply index

Thread overview: 74+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-08-20 12:04 [PATCH v5 00/31] Add FADump support on PowerNV platform Hari Bathini
2019-08-20 12:04 ` [PATCH v5 01/31] powerpc/fadump: move internal macros/definitions to a new header Hari Bathini
2019-09-03 11:09   ` Michael Ellerman
2019-09-03 16:05     ` Hari Bathini
2019-08-20 12:04 ` [PATCH v5 02/31] powerpc/fadump: move internal code to a new file Hari Bathini
2019-09-03 11:09   ` Michael Ellerman
2019-09-03 16:05     ` Hari Bathini
2019-09-04  9:02       ` Mahesh Jagannath Salgaonkar
2019-09-04 18:26         ` Hari Bathini
2019-08-20 12:04 ` [PATCH v5 03/31] powerpc/fadump: Improve fadump documentation Hari Bathini
2019-08-20 12:04 ` [PATCH v5 04/31] pseries/fadump: move rtas specific definitions to platform code Hari Bathini
2019-08-20 12:04 ` [PATCH v5 05/31] pseries/fadump: introduce callbacks for platform specific operations Hari Bathini
2019-09-03 11:10   ` Michael Ellerman
2019-09-03 16:06     ` Hari Bathini
2019-09-06  6:39       ` Hari Bathini
2019-08-20 12:04 ` [PATCH v5 06/31] pseries/fadump: define register/un-register callback functions Hari Bathini
2019-09-03 11:10   ` Michael Ellerman
2019-09-03 17:15     ` Hari Bathini
2019-08-20 12:04 ` [PATCH v5 07/31] powerpc/fadump: release all the memory above boot memory size Hari Bathini
2019-09-03 11:10   ` Michael Ellerman
2019-09-03 16:27     ` Hari Bathini
2019-08-20 12:05 ` [PATCH v5 08/31] pseries/fadump: move out platform specific support from generic code Hari Bathini
2019-08-20 12:05 ` [PATCH v5 09/31] powerpc/fadump: use FADump instead of fadump for how it is pronounced Hari Bathini
2019-08-20 12:05 ` [PATCH v5 10/31] opal: add MPIPL interface definitions Hari Bathini
2019-09-03 11:10   ` Michael Ellerman
2019-09-03 16:28     ` Hari Bathini
2019-09-04 11:03       ` Michael Ellerman
2019-09-04 11:05   ` Michael Ellerman
2019-08-20 12:05 ` [PATCH v5 11/31] powernv/fadump: add fadump support on powernv Hari Bathini
2019-09-03 11:10   ` Michael Ellerman
2019-09-03 16:31     ` Hari Bathini
2019-09-04 14:33       ` Hari Bathini
2019-09-05  3:11         ` Michael Ellerman
2019-08-20 12:05 ` [PATCH v5 12/31] powernv/fadump: register kernel metadata address with opal Hari Bathini
2019-09-04 11:25   ` Michael Ellerman
2019-08-20 12:05 ` [PATCH v5 13/31] powernv/fadump: reset metadata address during clean up Hari Bathini
2019-08-27 12:00   ` Hari Bathini
2019-08-20 12:05 ` [PATCH v5 14/31] powernv/fadump: define register/un-register callback functions Hari Bathini
2019-09-05  4:15   ` Michael Ellerman
2019-09-05  7:23   ` Michael Ellerman
2019-09-05  9:54     ` Hari Bathini
2019-08-20 12:05 ` [PATCH v5 15/31] powernv/fadump: support copying multiple kernel boot memory regions Hari Bathini
2019-09-04 11:30   ` Michael Ellerman
2019-09-04 20:20     ` Hari Bathini
2019-09-05  3:13       ` Michael Ellerman
2019-08-20 12:06 ` [PATCH v5 16/31] powernv/fadump: process the crashdump by exporting it as /proc/vmcore Hari Bathini
2019-09-04 11:42   ` Michael Ellerman
2019-09-04 21:01     ` Hari Bathini
2019-08-20 12:06 ` [PATCH v5 17/31] powernv/fadump: Warn before processing partial crashdump Hari Bathini
2019-09-04 11:48   ` Michael Ellerman
2019-08-20 12:06 ` [PATCH v5 18/31] powernv/fadump: handle invalidation of crashdump and re-registraion Hari Bathini
2019-08-20 12:06 ` [PATCH v5 19/31] powerpc/fadump: Update documentation about OPAL platform support Hari Bathini
2019-09-04 11:51   ` Michael Ellerman
2019-09-04 12:08     ` Oliver O'Halloran
2019-09-05  3:15       ` Michael Ellerman
2019-08-20 12:06 ` [PATCH v5 20/31] powerpc/fadump: use smaller offset while finding memory for reservation Hari Bathini
2019-09-04 11:54   ` Michael Ellerman
2019-08-20 12:06 ` [PATCH v5 21/31] powernv/fadump: process architected register state data provided by firmware Hari Bathini
2019-09-04 12:20   ` Michael Ellerman
2019-09-09 13:23     ` Hari Bathini
2019-09-09 15:33       ` Oliver O'Halloran
2019-09-10  8:48         ` Hari Bathini
2019-09-10 14:05           ` Michael Ellerman
2019-09-10 16:10             ` Hari Bathini [this message]
2019-08-20 12:06 ` [PATCH v5 22/31] powerpc/fadump: make crash memory ranges array allocation generic Hari Bathini
2019-08-20 12:06 ` [PATCH v5 23/31] powerpc/fadump: consider reserved ranges while releasing memory Hari Bathini
2019-08-20 12:07 ` [PATCH v5 24/31] powerpc/fadump: improve how crashed kernel's memory is reserved Hari Bathini
2019-08-20 12:07 ` [PATCH v5 25/31] powernv/fadump: add support to preserve crash data on FADUMP disabled kernel Hari Bathini
2019-08-20 12:07 ` [PATCH v5 26/31] powerpc/fadump: update documentation about CONFIG_PRESERVE_FA_DUMP Hari Bathini
2019-08-20 12:07 ` [PATCH v5 27/31] powernv/opalcore: export /sys/firmware/opal/core for analysing opal crashes Hari Bathini
2019-08-20 12:07 ` [PATCH v5 28/31] powernv/opalcore: provide an option to invalidate /sys/firmware/opal/core file Hari Bathini
2019-08-20 12:07 ` [PATCH v5 29/31] powerpc/fadump: consider f/w load area Hari Bathini
2019-08-20 12:07 ` [PATCH v5 30/31] powernv/fadump: update documentation about option to release opalcore Hari Bathini
2019-08-20 12:07 ` [PATCH v5 31/31] powernv/fadump: support holes in kernel boot memory area Hari Bathini

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=8b26238d-d8f6-c9a5-714e-b4682b94f8a5@linux.ibm.com \
    --to=hbathini@linux.ibm.com \
    --cc=ananth@linux.ibm.com \
    --cc=dja@axtens.net \
    --cc=hegdevasant@linux.ibm.com \
    --cc=linuxppc-dev@ozlabs.org \
    --cc=mahesh@linux.ibm.com \
    --cc=mpe@ellerman.id.au \
    --cc=npiggin@gmail.com \
    --cc=oohall@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

LinuxPPC-Dev Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linuxppc-dev/0 linuxppc-dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linuxppc-dev linuxppc-dev/ https://lore.kernel.org/linuxppc-dev \
		linuxppc-dev@lists.ozlabs.org linuxppc-dev@ozlabs.org linuxppc-dev@archiver.kernel.org
	public-inbox-index linuxppc-dev

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.ozlabs.lists.linuxppc-dev


AGPL code for this site: git clone https://public-inbox.org/ public-inbox