From: Ian Campbell <ijc@hellion.org.uk>
To: huang ying <huang.ying.caritas@gmail.com>
Cc: "Huang, Ying" <ying.huang@intel.com>,
akpm@linux-foundation.org, "H. Peter Anvin" <hpa@zytor.com>,
Thomas Gleixner <tglx@linutronix.de>,
Ingo Molnar <mingo@redhat.com>, Andi Kleen <ak@suse.de>,
linux-kernel@vger.kernel.org,
Jeremy Fitzhardinge <jeremy@goop.org>
Subject: Re: [PATCH -mm 1/3] i386 boot: replace boot_ioremap with enhanced bt_ioremap - enhance bt_ioremap
Date: Fri, 18 Jan 2008 15:23:57 +0000 [thread overview]
Message-ID: <1200669837.9891.35.camel@localhost.localdomain> (raw)
In-Reply-To: <851fc09e0801180654u6957f966g47c7629db6d4b2b0@mail.gmail.com>
On Fri, 2008-01-18 at 22:54 +0800, huang ying wrote:
> On Jan 18, 2008 4:48 PM, Ian Campbell <ijc@hellion.org.uk> wrote:
> > On Tue, 2008-01-15 at 13:45 +0800, Huang, Ying wrote:
> > > +void __init bt_ioremap_init(void)
> > > +{
> > > [...]
> > > + *pgd = __pa(bm_pte) | _PAGE_TABLE;
> > > +}
> > > [...]
> > > +static void __init __bt_set_fixmap(enum fixed_addresses idx,
> > > + unsigned long phys, pgprot_t flags)
> > > +{
> > > [...]
> > > + if (pgprot_val(flags))
> > > + *pte = (phys & PAGE_MASK) | pgprot_val(flags);
> > > + else
> > > + *pte = 0;
> > [...]
> >
> > Shouldn't these, and the rest of the file, be using the PTE accessor
> > macros set_pte,clear_pte etc? The boot_ioremap it replaces seems to have
> > done. Otherwise these patches don't appear to be paravirt_ops clean. I
>
> If CONFIG_X86_PAE is defined, the set_pte, clear_pte etc will operate
> 3-level page tables, while on i386, the early page table is always
> 2-level, so set_pte, clear_pte etc functions can not be used here. The
> boot_ioremap use a trick to deal with this problem. The CONFIG_X86_PAE
> is undefined in arch/x86/mm/boot_ioremap_32.c unconditionally, so the
> 2-level page table handling function is always used.
Ah, when booting a Xen guest you get 3 level page tables right from the
start.
I'm hacking on a patch right now but I hadn't appreciated/noticed that
this was a two level page table even on a PAE kernel which would explain
why it isn't exactly working as planned...
> Is the method used by boot_ioremap better for Xen?
Well, it worked ;-) I haven't looked closely enough at the two ways of
doing things to comment further but I'm sure the new way can be made to
work for Xen too somehow.
> [snip]
> This crash has nothing to do with this patch. Because this patch is
> for i386 only, x86_64 has its own early_ioremap implementation.
It's a 64 bit hypervisor but the guest in question is a 32 bit (PAE)
guest.
Ian
--
Ian Campbell
Current Noise: Mistress - Cheyne Stoking
You can't kiss a girl unexpectedly -- only sooner than she thought you would.
prev parent reply other threads:[~2008-01-18 15:24 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2008-01-15 5:45 Huang, Ying
2008-01-18 8:48 ` Ian Campbell
2008-01-18 12:50 ` Ingo Molnar
2008-01-18 14:54 ` huang ying
2008-01-18 14:54 ` H. Peter Anvin
2008-01-18 15:54 ` Ian Campbell
2008-01-18 16:22 ` Ingo Molnar
2008-01-18 17:27 ` Jeremy Fitzhardinge
2008-01-18 17:45 ` Ian Campbell
2008-01-18 15:23 ` Ian Campbell [this message]
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=1200669837.9891.35.camel@localhost.localdomain \
--to=ijc@hellion.org.uk \
--cc=ak@suse.de \
--cc=akpm@linux-foundation.org \
--cc=hpa@zytor.com \
--cc=huang.ying.caritas@gmail.com \
--cc=jeremy@goop.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=tglx@linutronix.de \
--cc=ying.huang@intel.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
Powered by JetHome