linux-kernel - Re: SNP guest crash in memblock with unaccepted memory

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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <aBCAuEJxBp9MPrOm@kernel.org>
Date: Tue, 29 Apr 2025 10:33:12 +0300
From: Mike Rapoport <rppt@...nel.org>
To: Tom Lendacky <thomas.lendacky@....com>
Cc: David Hildenbrand <david@...hat.com>,
	"Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Vlastimil Babka <vbabka@...e.cz>,
	Mel Gorman <mgorman@...hsingularity.net>,
	"Kalra, Ashish" <ashish.kalra@....com>,
	Rick Edgecombe <rick.p.edgecombe@...el.com>, linux-mm@...ck.org,
	linux-coco@...ts.linux.dev, linux-kernel@...r.kernel.org,
	Michael Roth <michael.roth@....com>
Subject: Re: SNP guest crash in memblock with unaccepted memory

On Mon, Apr 28, 2025 at 01:10:31PM -0500, Tom Lendacky wrote:
> On 4/28/25 09:04, David Hildenbrand wrote:
> > On 27.04.25 17:01, Tom Lendacky wrote:
> >> Hi Kirill,
> >>
> >> Every now and then I experience an SNP guest boot failure for accessing
> >> memory that hasn't been accepted. I managed to get a back trace:
> >>
> >>    RIP: 0010:memcpy_orig+0x68/0x130
> >>    Code: ...
> >>    RSP: 0000:ffffffff9cc03ce8 EFLAGS: 00010006
> >>    RAX: ff11001ff83e5000 RBX: 0000000000000000 RCX: fffffffffffff000
> >>    RDX: 0000000000000bc0 RSI: ffffffff9dba8860 RDI: ff11001ff83e5c00
> >>    RBP: 0000000000002000 R08: 0000000000000000 R09: 0000000000002000
> >>    R10: 000000207fffe000 R11: 0000040000000000 R12: ffffffff9d06ef78
> >>    R13: ff11001ff83e5000 R14: ffffffff9dba7c60 R15: 0000000000000c00
> >>    memblock_double_array+0xff/0x310
> >>    memblock_add_range+0x1fb/0x2f0
> >>    memblock_reserve+0x4f/0xa0
> >>    memblock_alloc_range_nid+0xac/0x130
> >>    memblock_alloc_internal+0x53/0xc0
> >>    memblock_alloc_try_nid+0x3d/0xa0
> >>    swiotlb_init_remap+0x149/0x2f0
> >>    mem_init+0xb/0xb0
> >>    mm_core_init+0x8f/0x350
> >>    start_kernel+0x17e/0x5d0
> >>    x86_64_start_reservations+0x14/0x30
> >>    x86_64_start_kernel+0x92/0xa0
> >>    secondary_startup_64_no_verify+0x194/0x19b
> >>
> >> I don't know a lot about memblock, but it appears that it needs to
> >> allocate more memory for it's regions array and returns a range of memory
> >> that hasn't been accepted. When the memcpy() runs, the SNP guest gets a
> >> #VC 0x404 because of this.
> >>
> >> Do you think it is as simple as calling accept_memory() on the memory
> >> range returned from memblock_find_in_range() in memblock_double_array()?
> > 
> > (not Kirill, but replying :) )
> > 
> > Yeah, we seem to be effectively allocating memory from memblock ("from
> > ourselves") without considering that memory must be accepted first.
> > 
> > accept_memory() on the new memory (in case of !slab) should be the right
> > thing to do.
> 
> Thanks, David. Let me add a call in for accept_memory in the !slab case
> and see if that resolves it. May take a bit to repro, but should find
> out eventually.

I think sticking a loop of memblock_alloc() somewhere before mm_core_init()
should trigger the issue.
 
> I'll submit a patch once I verify.
> 
> Thanks,
> Tom
> 
> > 

-- 
Sincerely yours,
Mike.