From: Chuck Zmudzinski <brchuckz@netscape.net>
To: Jan Beulich <jbeulich@suse.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>,
Andy Lutomirski <luto@kernel.org>,
Peter Zijlstra <peterz@infradead.org>,
Thomas Gleixner <tglx@linutronix.de>,
Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
Dan Williams <dan.j.williams@intel.com>,
Tom Lendacky <thomas.lendacky@amd.com>,
Jane Chu <jane.chu@oracle.com>,
Tianyu Lan <Tianyu.Lan@microsoft.com>,
Randy Dunlap <rdunlap@infradead.org>,
Sean Christopherson <seanjc@google.com>,
xen-devel@lists.xenproject.org, stable@vger.kernel.org,
linux-kernel@vger.kernel.org, Juergen Gross <jgross@suse.com>,
"Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Subject: Re: [PATCH v2] Subject: x86/PAT: Report PAT on CPUs that support PAT without MTRR
Date: Wed, 13 Jul 2022 11:02:06 -0400 [thread overview]
Message-ID: <dc0ee2d8-fc88-e4f1-6867-43d3ca3732b2@netscape.net> (raw)
In-Reply-To: <dbfd3a14-781e-c66e-b11c-e21ba4134067@suse.com>
On 7/13/2022 9:52 AM, Jan Beulich wrote:
> On 13.07.2022 15:49, Chuck Zmudzinski wrote:
> > On 7/13/2022 9:34 AM, Jan Beulich wrote:
> >> On 13.07.2022 13:10, Chuck Zmudzinski wrote:
> >>> On 7/13/2022 6:36 AM, Chuck Zmudzinski wrote:
> >>>> On 7/13/2022 5:09 AM, Jan Beulich wrote:
> >>>>> On 13.07.2022 10:51, Chuck Zmudzinski wrote:
> >>>>>> On 7/13/22 2:18 AM, Jan Beulich wrote:
> >>>>>>> On 13.07.2022 03:36, Chuck Zmudzinski wrote:
> >>>>>>>> v2: *Add force_pat_disabled variable to fix "nopat" on Xen PV (Jan Beulich)
> >>>>>>>> *Add the necessary code to incorporate the "nopat" fix
> >>>>>>>> *void init_cache_modes(void) -> void __init init_cache_modes(void)
> >>>>>>>> *Add Jan Beulich as Co-developer (Jan has not signed off yet)
> >>>>>>>> *Expand the commit message to include relevant parts of the commit
> >>>>>>>> message of Jan Beulich's proposed patch for this problem
> >>>>>>>> *Fix 'else if ... {' placement and indentation
> >>>>>>>> *Remove indication the backport to stable branches is only back to 5.17.y
> >>>>>>>>
> >>>>>>>> I think these changes address all the comments on the original patch
> >>>>>>>>
> >>>>>>>> I added Jan Beulich as a Co-developer because Juergen Gross asked me to
> >>>>>>>> include Jan's idea for fixing "nopat" that was missing from the first
> >>>>>>>> version of the patch.
> >>>>>>>
> >>>>>>> You've sufficiently altered this change to clearly no longer want my
> >>>>>>> S-o-b; unfortunately in fact I think you broke things:
> >>>>>>
> >>>>>> Well, I hope we can come to an agreement so I have
> >>>>>> your S-o-b. But that would probably require me to remove
> >>>>>> Juergen's R-b.
> >>>>>>
> >>>>>>>> @@ -292,7 +294,7 @@ void init_cache_modes(void)
> >>>>>>>> rdmsrl(MSR_IA32_CR_PAT, pat);
> >>>>>>>> }
> >>>>>>>>
> >>>>>>>> - if (!pat) {
> >>>>>>>> + if (!pat || pat_force_disabled) {
> >>>>>>>
> >>>>>>> By checking the new variable here ...
> >>>>>>>
> >>>>>>>> /*
> >>>>>>>> * No PAT. Emulate the PAT table that corresponds to the two
> >>>>>>>> * cache bits, PWT (Write Through) and PCD (Cache Disable).
> >>>>>>>> @@ -313,6 +315,16 @@ void init_cache_modes(void)
> >>>>>>>> */
> >>>>>>>> pat = PAT(0, WB) | PAT(1, WT) | PAT(2, UC_MINUS) | PAT(3, UC) |
> >>>>>>>> PAT(4, WB) | PAT(5, WT) | PAT(6, UC_MINUS) | PAT(7, UC);
> >>>>>>>
> >>>>>>> ... you put in place a software view which doesn't match hardware. I
> >>>>>>> continue to think that ...
> >>>>>>>
> >>>>>>>> + } else if (!pat_bp_enabled) {
> >>>>>>>
> >>>>>>> ... the variable wants checking here instead (at which point, yes,
> >>>>>>> this comes quite close to simply being a v2 of my original patch).
> >>>>>>>
> >>>>>>> By using !pat_bp_enabled here you actually broaden where the change
> >>>>>>> would take effect. Iirc Boris had asked to narrow things (besides
> >>>>>>> voicing opposition to this approach altogether). Even without that
> >>>>>>> request I wonder whether you aren't going to far with this.
> >>>>>>>
> >>>>>>> Jan
> >>>>>>
> >>>>>> I thought about checking for the administrator's "nopat"
> >>>>>> setting where you suggest which would limit the effect
> >>>>>> of "nopat" to not reporting PAT as enabled to device
> >>>>>> drivers who query for PAT availability using pat_enabled().
> >>>>>> The main reason I did not do that is that due to the fact
> >>>>>> that we cannot write to the PAT MSR, we cannot really
> >>>>>> disable PAT. But we come closer to respecting the wishes
> >>>>>> of the administrator by configuring the caching modes as
> >>>>>> if PAT is actually disabled by the hardware or firmware
> >>>>>> when in fact it is not.
> >>>>>>
> >>>>>> What would you propose logging as a message when
> >>>>>> we report PAT as disabled via pat_enabled()? The main
> >>>>>> reason I did not choose to check the new variable in the
> >>>>>> new 'else if' block is that I could not figure out what to
> >>>>>> tell the administrator in that case. I think we would have
> >>>>>> to log something like, "nopat is set, but we cannot disable
> >>>>>> PAT, doing our best to disable PAT by not reporting PAT
> >>>>>> as enabled via pat_enabled(), but that does not guarantee
> >>>>>> that kernel drivers and components cannot use PAT if they
> >>>>>> query for PAT support using boot_cpu_has(X86_FEATURE_PAT)
> >>>>>> instead of pat_enabled()." However, I acknowledge WC mappings
> >>>>>> would still be disabled because arch_can_pci_mmap_wc() will
> >>>>>> be false if pat_enabled() is false.
> >>>>>>
> >>>>>> Perhaps we also need to log something if we keep the
> >>>>>> check for "nopat" where I placed it. We could say something
> >>>>>> like: "nopat is set, but we cannot disable hardware/firmware
> >>>>>> PAT support, so we are emulating as if there is no PAT support
> >>>>>> which puts in place a software view that does not match
> >>>>>> hardware."
> >>>>>>
> >>>>>> No matter what, because we cannot write to PAT MSR in
> >>>>>> the Xen PV case, we probably need to log something to
> >>>>>> explain the problems associated with trying to honor the
> >>>>>> administrator's request. Also, what log level should it be.
> >>>>>> Should it be a pr_warn instead of a pr_info?
> >>>>>
> >>>>> I'm afraid I'm the wrong one to answer logging questions. As you
> >>>>> can see from my original patch, I didn't add any new logging (and
> >>>>> no addition was requested in the comments that I have got). I also
> >>>>> don't think "nopat" has ever meant "disable PAT", as the feature
> >>>>> is either there or not. Instead I think it was always seen as
> >>>>> "disable fiddling with PAT", which by implication means using
> >>>>> whatever is there (if the feature / MSR itself is available).
> >>>>
> >>>> IIRC, I do think I mentioned in the comments on your patch that
> >>>> it would be preferable to mention in the commit message that
> >>>> your patch would change the current behavior of "nopat" on
> >>>> Xen. The question is, how much do we want to change the
> >>>> current behavior of "nopat" on Xen. I think if we have to change
> >>>> the current behavior of "nopat" on Xen and if we are going
> >>>> to propagate that change to all current stable branches all
> >>>> the way back to 4.9.y,, we better make a lot of noise about
> >>>> what we are doing here.
> >>>>
> >>>> Chuck
> >>>
> >>> And in addition, if we are going to backport this patch to
> >>> all current stable branches, we better have a really, really,
> >>> good reason for changing the behavior of "nopat" on Xen.
> >>>
> >>> Does such a reason exist?
> >>
> >> Well, the simple reason is: It doesn't work the same way under Xen
> >> and non-Xen (in turn because, before my patch or whatever equivalent
> >> work, things don't work properly anyway, PAT-wise). Yet it definitely
> >> ought to behave the same everywhere, imo.
> >>
> >> Jan
> >
> > IOW, you are saying PAT has been broken on Xen for a
> > long time, and it is necessary to fix it now not only on
> > master, but also on all the stable branches.
> >
> > Why is it necessary to do it on all the stable branches?
>
> I'm not saying that's _necessary_ (but I think it would make sense),
> and I'm not the one to decide whether or how far to backport this.
>
> Jan
What conclusion do you draw from these facts?
1. Linus' regression rule is a high priority in Linux
2. Security concerns are even a higher priority in Linux
3. You and I have been trying to fix a regression for the past two months
4. The ones who can fix the regression have not accepted our patches.
5. I have been asked to help backport my fix to all stable branches.
Chuck
next prev parent reply other threads:[~2022-07-13 15:02 UTC|newest]
Thread overview: 37+ messages / expand[flat|nested] mbox.gz Atom feed top
[not found] <9d5070ae4f3e956a95d3f50e24f1a93488b9ff52.1657671676.git.brchuckz.ref@aol.com>
2022-07-13 1:36 ` [PATCH v2] Subject: x86/PAT: Report PAT on CPUs that support PAT without MTRR Chuck Zmudzinski
2022-07-13 4:12 ` Juergen Gross
2022-07-13 6:18 ` Jan Beulich
2022-07-13 8:51 ` Chuck Zmudzinski
2022-07-13 9:09 ` Jan Beulich
2022-07-13 10:36 ` Chuck Zmudzinski
2022-07-13 11:10 ` Chuck Zmudzinski
2022-07-13 13:34 ` Jan Beulich
2022-07-13 13:45 ` Juergen Gross
2022-07-13 17:01 ` Chuck Zmudzinski
2022-07-13 19:07 ` Chuck Zmudzinski
2022-07-13 19:22 ` Chuck Zmudzinski
2022-07-13 19:38 ` Chuck Zmudzinski
2022-07-13 22:12 ` Chuck Zmudzinski
2022-07-13 13:49 ` Chuck Zmudzinski
2022-07-13 13:52 ` Jan Beulich
2022-07-13 15:02 ` Chuck Zmudzinski [this message]
2022-07-13 15:05 ` Jan Beulich
2022-07-14 5:30 ` Thorsten Leemhuis
2022-07-15 2:07 ` Chuck Zmudzinski
2022-07-15 5:00 ` Thorsten Leemhuis
2022-07-15 10:05 ` Jan Beulich
2022-07-15 19:53 ` Chuck Zmudzinski
2022-07-16 11:02 ` Chuck Zmudzinski
2022-07-18 6:07 ` Jan Beulich
2022-07-18 11:31 ` Chuck Zmudzinski
2022-07-18 11:39 ` Jan Beulich
2022-07-18 11:45 ` Chuck Zmudzinski
2022-07-18 12:12 ` Chuck Zmudzinski
2022-08-17 16:39 ` Chuck Zmudzinski
2022-07-14 5:40 ` Juergen Gross
2022-07-14 6:28 ` Jan Beulich
2022-07-15 2:19 ` Chuck Zmudzinski
2022-07-15 2:53 ` Chuck Zmudzinski
2022-07-15 2:58 ` Chuck Zmudzinski
2022-07-15 4:22 ` Juergen Gross
2022-07-15 4:42 ` Chuck Zmudzinski
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=dc0ee2d8-fc88-e4f1-6867-43d3ca3732b2@netscape.net \
--to=brchuckz@netscape.net \
--cc=Tianyu.Lan@microsoft.com \
--cc=bp@alien8.de \
--cc=dan.j.williams@intel.com \
--cc=dave.hansen@linux.intel.com \
--cc=hpa@zytor.com \
--cc=jane.chu@oracle.com \
--cc=jbeulich@suse.com \
--cc=jgross@suse.com \
--cc=kirill.shutemov@linux.intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=luto@kernel.org \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=rdunlap@infradead.org \
--cc=seanjc@google.com \
--cc=stable@vger.kernel.org \
--cc=tglx@linutronix.de \
--cc=thomas.lendacky@amd.com \
--cc=x86@kernel.org \
--cc=xen-devel@lists.xenproject.org \
/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;
as well as URLs for NNTP newsgroup(s).