Message-ID: <20190403082305.GG1555@localhost.localdomain>
Date:   Wed, 3 Apr 2019 16:23:06 +0800
From:   Chao Fan <fanc.fnst@...fujitsu.com>
To:     Dave Young <dyoung@...hat.com>
CC:     Junichi Nomura <j-nomura@...jp.nec.com>,
        Borislav Petkov <bp@...en8.de>,
        "bhe@...hat.com" <bhe@...hat.com>,
        "kasong@...hat.com" <kasong@...hat.com>,
        "x86@...nel.org" <x86@...nel.org>,
        "kexec@...ts.infradead.org" <kexec@...ts.infradead.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2] x86/boot: Use efi_setup_data for searching RSDP on
 kexec-ed kernel

On Wed, Apr 03, 2019 at 04:09:16PM +0800, Dave Young wrote:
>On 04/03/19 at 03:21pm, Chao Fan wrote:
>> On Wed, Apr 03, 2019 at 01:53:40PM +0800, Dave Young wrote:
>> >On 04/03/19 at 01:35pm, Chao Fan wrote:
>> >> On Tue, Apr 02, 2019 at 08:03:19PM +0800, Dave Young wrote:
>> >> >On 04/01/19 at 12:08am, Junichi Nomura wrote:
>> >> >> Commit 3a63f70bf4c3a ("x86/boot: Early parse RSDP and save it in
>> >> >> boot_params") broke kexec boot on EFI systems.  efi_get_rsdp_addr()
>> >> >> in the early parsing code tries to search RSDP from EFI table but
>> >> >> that will crash because the table address is virtual when the kernel
>> >> >> was booted by kexec.
>> >> >> 
>> >> >> In the case of kexec, physical address of EFI tables is provided
>> >> >> via efi_setup_data in boot_params, which is set up by kexec(1).
>> >> >> 
>> >> >> Factor out the table parsing code and use different pointers depending
>> >> >> on whether the kernel is booted by kexec or not.
>> >> >> 
>> >> >> Fixes: 3a63f70bf4c3a ("x86/boot: Early parse RSDP and save it in boot_params")
>> >> >> Signed-off-by: Jun'ichi Nomura <j-nomura@...jp.nec.com>
>> >> >> Acked-by: Baoquan He <bhe@...hat.com>
>> >> >> Cc: Chao Fan <fanc.fnst@...fujitsu.com>
>> >> >> Cc: Borislav Petkov <bp@...e.de>
>> >> >> Cc: Dave Young <dyoung@...hat.com>
>> >> [...]
>> >> >
>> >> >I failed to kexec reboot on my laptop, kernel panics too quick,  I'm not sure this is
>> >> >caused by your patch though.
>> >> >
>> >> >Actually there are something probably i915 changes break kexec,  the
>> >> >above test is with "nomodeset" which should work.
>> >> >
>> >> >Let me do more testing and update here tomorrow.
>> >> >
>> >> 
>> >> Hi Dave,
>> >> 
>> >> Last day I was testing the normal kexec, today I have tested the kdump
>> >> issue. Since the kdump has set "nokaslr" to cmdline, so I drop from
>> >> KDUMP_COMMANDLINE_APPEND
>> >> And it booted OK, so the PATCH works in both normal kexec and kdump.
>> >> 
>> >
>> >Actually I got some different kexec test results.
>> >
>> >Yesterday, with my installed kernel (based on git head several weeks
>> >ago), kexec kernel panics.
>> >
>> >Then I tried latest mainline with git pull, everything works, (with or
>> >without the patch, and can not reproduce the bug this patch is fixing)
>> >
>> >Today, test again, kexec reboot hangs (with or without your patch), but
>> >kdump works always (with or without the patch)
>> 
>> That's so strange, I'm puzzled by your test.
>> Withour ant changes, just test again, the result of yesterday and today
>> are different?
>> 
>> >
>> >It is weird to me. Probably I need find out why I can not reproduce the
>> >bug this patch is addressing first.
>> 
>> Hmm, is your laptop booted from EFI or legacy BIOS?
>
>EFI booted.
>
>> >
>> >earlyprintk seems not working for me anymore, it is not easy to debug on
>> >laptop now.
>> 
>> Another computer can help print the console message.
>> I remeber you are good at debug on two computers.
>
>I used to use earlyprintk, but seems earlyprintk can not work so early.
>So I only can try to comment out code manually.

Ah, I got it. If you want to use earlyprintk to print message of
get_rsdp_addr(), you should change some code.
Because get_rsdp_addr() is called before earlyprintk initialed.

I am still testing, but in my debug stage, I change like this:
diff --git a/arch/x86/boot/compressed/misc.c b/arch/x86/boot/compressed/misc.c
index c0d6c560df69..dbbe8d9a5792 100644
--- a/arch/x86/boot/compressed/misc.c
+++ b/arch/x86/boot/compressed/misc.c
@@ -352,7 +352,6 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,
        boot_params->hdr.loadflags &= ~KASLR_FLAG;

        /* Save RSDP address for later use. */
-       boot_params->acpi_rsdp_addr = get_rsdp_addr();

        sanitize_boot_params(boot_params);

@@ -368,6 +367,7 @@ asmlinkage __visible void *extract_kernel(void *rmode, memptr heap,
        cols = boot_params->screen_info.orig_video_cols;

        console_init();
+       boot_params->acpi_rsdp_addr = get_rsdp_addr();
        debug_putstr("early console in extract_kernel\n");

        free_mem_ptr     = heap;        /* Heap */

Call this function after console_init().
That will make earlyprintk works.

>
>It seems to be still some bug in early acpi code because if I 
>return 0 in get rsdp function it just works.
>
>This guess make sense to me for now:
>1. nokaslr does not work because it may only avoid randomization, but it
>does not avoid running the early acpi parsing code.
>
>2. kexec does not work without the patch we talked about, because I just
>reproduced Junichi reported bug, sometimes hang, or reset, it is
>understandable.
>
>3. kexec does not work with the patch, so there is still bugs in the
>code. 
>
>So several things to do is:
>
>Fix 1. (separate issue),  Baoquan or either of you may want to fix it :)

For 1, I think it's not needed to fix. Since not only KASLR needs RSDP.
So nokaslr should not stop the acpi parsing code.
>
>Fix 3.  need more debugging, have you or Junichi run tests on more real
>hardware, maybe it is easier to reproduce on real hardware, I'm glad to
>help to try test patch or provide any help. 

I am still testing in real hardware.

Thanks,
Chao Fan

>
>Thanks
>Dave
>
>

Powered by blists - more mailing lists