mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Dmitry Osipenko <digetx@gmail.com>
To: Mikko Perttunen <cyndis@kapsi.fi>,
	Mikko Perttunen <mperttunen@nvidia.com>,
	thierry.reding@gmail.com, jonathanh@nvidia.com
Cc: dri-devel@lists.freedesktop.org, linux-tegra@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 1/4] gpu: host1x: Enable Tegra186 syncpoint protection
Date: Sat, 19 Aug 2017 14:51:37 +0300	[thread overview]
Message-ID: <605615e6-2981-1443-03ad-ed37177b017c@gmail.com> (raw)
In-Reply-To: <72fc513b-2626-fd86-1b98-5d2f8e89dc5f@kapsi.fi>

On 19.08.2017 14:32, Mikko Perttunen wrote:
> 
> 
> On 08/19/2017 02:11 PM, Dmitry Osipenko wrote:
>> On 19.08.2017 13:35, Mikko Perttunen wrote:
>>> On 08/19/2017 01:09 PM, Dmitry Osipenko wrote:
>>>> On 19.08.2017 11:10, Mikko Perttunen wrote:
>>>> [snip]
>>>>>>> +    host1x_hw_syncpt_set_protection(host, true);
>>>>>>
>>>>>> Is it really okay to force the protection? Maybe protection should be enabled
>>>>>> with a respect to CONFIG_TEGRA_HOST1X_FIREWALL? In that case we would have to
>>>>>> avoid software jobs validation for Tegra124+.
>>>>>
>>>>> I don't quite get your comment. The hardware syncpt protection layer being
>>>>> enabled should never hurt - it doesn't mess with any valid jobs. It's also
>>>>> only
>>>>> on Tegra186 so I'm not sure where the Tegra124 comes from.
>>>>
>>>> Right, it's the gather filter on T124+, my bad. This raises several questions.
>>>>
>>>> 1) Why we have CONFIG_TEGRA_HOST1X_FIREWALL? Should it be always enforced or we
>>>> actually want to be a bit more flexible and allow to disable it. Imagine that
>>>> you are making a custom application and want to utilize channels in a
>>>> different way.
>>>
>>> I think it should be up to the user to decide whether they want the firewall or
>>> not. It's clearly the most useful on the older chips - especially Tegra20 due to
>>> lack of IOMMU. The performance penalty is too great to force it on always.
>>>
>>
>> Of course there is some overhead but is not that great. Usually command buffer
>> contains just a dozen of commands. It should be an interesting challenge to
>> optimize its performance though.
>>
>>> The programming model should always be considered the same - the rules of what
>>> you are allowed to do are the same whether the firewall, or any
>>> hardware-implemented protection features, are on or not.
>>>
>>
>> Well, okay.
>>
>>>>
>>>> 2) Since syncpoint protection is a T186 feature, what about previous
>>>> generations? Should we validate syncpoints in software for them? We have
>>>> 'syncpoint validation' patch staged in grate's kernel
>>>> https://github.com/grate-driver/linux/commit/c8b6c82173f2ee9fead23380e8330b8099e7d5e7
>>>>
>>>>
>>>> (I'll start sending out this and other patches after a bit more thorough
>>>> testing.) Improperly used syncpoints potentially could allow one program to
>>>> damage others.
>>>
>>> Yes, I think the firewall should have this feature for older generations. We
>>> could disable the check on Tegra186, as you point towards in question 4.
>>>
>>>>
>>>> 3) What exactly does gather filter? Could you list all the commands that it
>>>> filters out, please?
>>>
>>> According to the Tegra186 TRM (section 16.8.32), SETCLASS, SETSTRMID and EXTEND
>>> are filtered.
>>>
>>
>> Okay, then what about SETSTRMID command, I don't see its disassembly in the
>> host1x gather debug dump. Is it accidentally missed?
>>
> 
> True, it's a new command in Tegra186 and I missed adding it to the disassembler.
> It's probably fine to add it in another patch since it's only intended for
> kernel use and it's useless without IOMMU support anyway (which we don't have
> currently on Tegra186).
> 

Yeah, but it probably would be more preferable that this patch would predate the
"gather filter" enabling.

>>>>
>>>> 4) What about T30/T114 that do not have gather filter? Should we validate those
>>>> commands for them in a software firewall?
>>>
>>> Yes, the firewall should validate that.
>>>
>>>>
>>>> So maybe we should implement several layers of validation in the SW firewall.
>>>> Like all layers for T20 (memory boundaries validation etc), software gather
>>>> filter for T30/114 and software syncpoint validation for T30/114/124/210.
>>>>
>>>
>>> That seems like a good idea.
>>
>> Alright, factoring out firewall from job.c probably should be the first step.
>>


-- 
Dmitry

  reply	other threads:[~2017-08-19 11:51 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-08-18 16:15 [PATCH 0/4] Miscellaneous improvements to Host1x and TegraDRM Mikko Perttunen
2017-08-18 16:15 ` [PATCH 1/4] gpu: host1x: Enable Tegra186 syncpoint protection Mikko Perttunen
2017-08-18 22:36   ` Dmitry Osipenko
2017-08-19  8:10     ` Mikko Perttunen
2017-08-19 10:09       ` Dmitry Osipenko
2017-08-19 10:35         ` Mikko Perttunen
2017-08-19 11:11           ` Dmitry Osipenko
2017-08-19 11:32             ` Mikko Perttunen
2017-08-19 11:51               ` Dmitry Osipenko [this message]
2017-08-19 12:02   ` Dmitry Osipenko
2017-08-20 16:18   ` Dmitry Osipenko
2017-08-20 16:59   ` Dmitry Osipenko
2017-08-20 18:13   ` Dmitry Osipenko
2017-08-18 16:15 ` [PATCH 2/4] gpu: host1x: Enable gather filter Mikko Perttunen
2017-08-19 10:42   ` Dmitry Osipenko
2017-08-19 10:46     ` Mikko Perttunen
2017-08-19 12:05       ` Dmitry Osipenko
2017-08-20 16:24   ` Dmitry Osipenko
2017-08-20 16:44     ` Dmitry Osipenko
2017-08-20 16:59       ` Dmitry Osipenko
2017-08-21 17:27         ` Mikko Perttunen
2017-08-21 17:28           ` Mikko Perttunen
2017-08-18 16:15 ` [PATCH 3/4] gpu: host1x: Improve debug disassembly formatting Mikko Perttunen
2017-08-18 21:54   ` Dmitry Osipenko
2017-08-18 16:15 ` [PATCH 4/4] drm/tegra: Use u64_to_user_ptr helper Mikko Perttunen
2017-08-18 22:05   ` Dmitry Osipenko
2017-08-19  8:06     ` Mikko Perttunen

Reply instructions:

You may reply publicly 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=605615e6-2981-1443-03ad-ed37177b017c@gmail.com \
    --to=digetx@gmail.com \
    --cc=cyndis@kapsi.fi \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=jonathanh@nvidia.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-tegra@vger.kernel.org \
    --cc=mperttunen@nvidia.com \
    --cc=thierry.reding@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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®