linux-kernel - Re: [RFC PATCH v2 00/51] 1G page support for guest

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] [day] [month] [year] [list]

Message-ID: <CAEvNRgFmq8DP_=V7mrY8qza3i9h4-Bn0OWt72iDj6mELu+BiZg@mail.gmail.com>
Date: Fri, 23 Jan 2026 16:07:57 -0800
From: Ackerley Tng <ackerleytng@...gle.com>
To: kvm@...r.kernel.org, linux-mm@...ck.org, linux-kernel@...r.kernel.org, 
	x86@...nel.org, linux-fsdevel@...r.kernel.org
Cc: david@...nel.org, fvdl@...gle.com, ira.weiny@...el.com, 
	jthoughton@...gle.com, michael.roth@....com, pankaj.gupta@....com, 
	rick.p.edgecombe@...el.com, seanjc@...gle.com, vannapurve@...gle.com, 
	yan.y.zhao@...el.com, "Kalyazin, Nikita" <kalyazin@...zon.co.uk>
Subject: Re: [RFC PATCH v2 00/51] 1G page support for guest_memfd

Ackerley Tng <ackerleytng@...gle.com> writes:

Re-using this thread to collect discussions related to guest_memfd
HugeTLB support, also trimmed cc list.

Here's the latest public version Vishal and I have:

  https://github.com/googleprodkernel/linux-cc/tree/wip-gmem-conversions-hugetlb-restructuring-12-08-25

On the guest_memfd call on 2026-01-22, Michael found another bug to do
with multiple threads trying to allocate within the same huge page at
the same time.

The fix we're using to make progress is to use hugetlb_fault_mutex_lock.

unsigned int gmem_hugetlb_mapping_index_lock(struct address_space *mapping,
					     pgoff_t index, u8 page_order)
{
	pgoff_t index_floor = round_down(index, 1ULL << page_order);

	return hugetlb_fault_mutex_lock(mapping, index_floor);
}

void gmem_hugetlb_mapping_index_unlock(unsigned int hash)
{
	hugetlb_fault_mutex_unlock(hash);
}

and then

static struct folio *kvm_gmem_get_folio(struct inode *inode, pgoff_t index)
{
        ... declarations ...

	if (gmem_is_hugetlb(gi->flags))
		lock_id = gmem_hugetlb_mapping_index_lock(mapping, index, gi->page_order);

        ... and this right at the end ...

	if (gmem_is_hugetlb(gi->flags))
		gmem_hugetlb_mapping_index_unlock(lock_id);
}

Yan also found some bugs (thanks!) and there's a discussion at [*].

[*] https://lore.kernel.org/all/CAEvNRgGG+xYhsz62foOrTeAxUCYxpCKCJnNgTAMYMV=w2eq+6Q@mail.gmail.com/