linux-kernel - Re: [PATCH 1/2] x86 boot: add E820_RESVD

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <86802c440806300215w31abbcb3h6a42395e11701713@mail.gmail.com>
Date:	Mon, 30 Jun 2008 02:15:41 -0700
From:	"Yinghai Lu" <yhlu.kernel@...il.com>
To:	"Huang, Ying" <ying.huang@...el.com>
Cc:	"H. Peter Anvin" <hpa@...or.com>, andi@...stfloor.org,
	mingo@...hat.com, tglx@...utronix.de, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 1/2] x86 boot: add E820_RESVD_KERN

On Mon, Jun 30, 2008 at 12:51 AM, Huang, Ying <ying.huang@...el.com> wrote:
> On Mon, 2008-06-30 at 00:34 -0700, Yinghai Lu wrote:
>> On Mon, Jun 30, 2008 at 12:03 AM, Huang, Ying <ying.huang@...el.com> wrote:
>> > On Fri, 2008-06-27 at 15:05 -0700, Yinghai Lu wrote:
>> >> On Thu, Jun 26, 2008 at 7:48 PM, Huang, Ying <ying.huang@...el.com> wrote:
>> >> > On Thu, 2008-06-26 at 19:22 -0700, Yinghai Lu wrote:
>> >> >> On Thu, Jun 26, 2008 at 2:47 AM, Yinghai Lu <yhlu.kernel@...il.com> wrote:
>> >> >> > On Thu, Jun 26, 2008 at 12:48 AM, Huang, Ying <ying.huang@...el.com> wrote:
>> >> >> >> On Thu, 2008-06-26 at 00:25 -0700, Yinghai Lu wrote:
>> >> >> >> [...]
>> >> >> >>> >                if (pfn >= limit_pfn)
>> >> >> >>> > @@ -977,7 +978,7 @@ u64 __init early_reserve_e820(u64 startt
>> >> >> >>> >                return 0;
>> >> >> >>> >
>> >> >> >>> >        addr = round_down(start + size - sizet, align);
>> >> >> >>> > -       e820_update_range(addr, sizet, E820_RAM, E820_RESERVED);
>> >> >> >>> > +       e820_update_range(addr, sizet, E820_RAM, E820_RESVD_KERN);
>> >> >> >>>
>> >> >> >>> this line is not needed.
>> >> >> >>
>> >> >> >> Why? Memory reserved by early_rserved_e820 should not be saved during
>> >> >> >> hibernation? shoudl not be saved by kdump?
>> >> >> >>
>> >> >
>> >> > Can you tell me why this line is not needed?
>> >> >
>> >> > [...]
>> >> >> some like the attach patch...
>> >> >>
>> >> >> you still can merge parse_setup_data parse_e820_ext
>> >> >> also entries in parse_e820_ext is not initialized..., __copy_e820_map
>> >> >> will do nothing.
>> >> >
>> >> > OK. Because some E820 entries are available after parse_setup_data(),
>> >> > it is better to call reserve_setup_data() after calling
>> >> > parse_setup_data() if update_e820_range() is used instead of
>> >> > reserve_early().
>> >>
>> >> please modify it and test on your platforms then submit to Ingo..
>> >
>> > It seems that there is an issue:
>> >
>> > - If parse_setup_data() is called before reserve_setup_data(), and there
>> > is a conflict between memory area used by setup_data and other memory
>> > area, it is possible that the contents of setup_data is changed. So that
>> > system may panic before reporting memory area conflict. And it seems
>> > that memory area conflict is not checked by e820_update_range().
>>
>> what is "other memory area"? returned from find_e820_area? no one use that yet.
>
> I mean memory area reserved with reserved_early() or e820_update_range()
> before reserve_setup_data() is called.

before parse_setup_data, reserve_early is called for
1. kernel text/data/bss + initial pgt
2. ramdisk
3. ebda
e820_update_range is not called.
at this time early_res have RAM reserved by kernel.

then setup_memory_map is called, so e820 have some ranges...directly
from e820 table

next need to call parse_setup_data
it will add some entries in e820

then reserve_setup_data is called, it will use e820_update_range to
reserve setup_data itself directly in e820

...

>
> And because there is no conflict check in e820_update_range(), what to
> deal with potential conflict between setup_data and other memory area
> regardless which one is reserved earlier?

find_e820_area will make sure it only find ram from e820 and it is not
conflict with early_res


>
>
> Another issue related:
>
> Because some memmap entries are available via extended E820 memmap
> (SETUP_E820_EXT), it is not strictly safe to use e820_update_range()
> between setup_memory_map() and parse_setup_data(). It may be better to
> parse extended E820 memmap right after setup_memory_map().

will look at the code again..

YH