From: Bibo Mao <maobibo@loongson.cn>
To: cuitao <cuitao@kylinos.cn>,
zhaotianrui@loongson.cn, chenhuacai@kernel.org,
loongarch@lists.linux.dev
Cc: kernel@xen0n.name
Subject: Re: [PATCH] LoongArch: KVM: Fix data race in kvm_ipi_regs_access
Date: Tue, 14 Apr 2026 09:50:33 +0800 [thread overview]
Message-ID: <ce3d170a-56a4-9343-49fa-2d8b3cbc9ebe@loongson.cn> (raw)
In-Reply-To: <20260414013316.306075-1-cuitao@kylinos.cn>
On 2026/4/14 上午9:33, cuitao wrote:
> Add spin_lock protection around ipi_state field accesses in
> kvm_ipi_regs_access(). get_user/put_user are performed
> outside the lock since they may sleep on page faults.
>
> Signed-off-by: cuitao <cuitao@kylinos.cn>
> ---
> arch/loongarch/kvm/intc/ipi.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/arch/loongarch/kvm/intc/ipi.c b/arch/loongarch/kvm/intc/ipi.c
> index 1f6ebbd0af5c..e6a4a461a0c8 100644
> --- a/arch/loongarch/kvm/intc/ipi.c
> +++ b/arch/loongarch/kvm/intc/ipi.c
> @@ -348,18 +348,26 @@ static int kvm_ipi_regs_access(struct kvm_device *dev,
> if (len == 4) {
> if (get_user(val, (uint32_t __user *)attr->addr))
> return -EFAULT;
> + spin_lock(&vcpu->arch.ipi_state.lock);
> *(uint32_t *)p = (uint32_t)val;
> + spin_unlock(&vcpu->arch.ipi_state.lock);
> } else if (len == 8) {
> if (get_user(val, (uint64_t __user *)attr->addr))
> return -EFAULT;
> + spin_lock(&vcpu->arch.ipi_state.lock);
> *(uint64_t *)p = val;
> + spin_unlock(&vcpu->arch.ipi_state.lock);
> }
> } else {
> if (len == 4) {
> + spin_lock(&vcpu->arch.ipi_state.lock);
> val = *(uint32_t *)p;
> + spin_unlock(&vcpu->arch.ipi_state.lock);
> return put_user(val, (uint32_t __user *)attr->addr);
> } else if (len == 8) {
> + spin_lock(&vcpu->arch.ipi_state.lock);
> val = *(uint64_t *)p;
> + spin_unlock(&vcpu->arch.ipi_state.lock);
> return put_user(val, (uint64_t __user *)attr->addr);
> }
> }
>
I am not sure whether spinlock need be added here. When get/set ipi
register of vCPU from VMM side, the vCPU should be kicked out at safe
state by VMM; Otherwise even if lock is added, VMM will get one register
at a time and vCPU may write these registers at the same time, part of
registers acquired by VMM still are stale.
Regards
Bibo Mao
next prev parent reply other threads:[~2026-04-14 1:53 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-04-14 1:33 [PATCH] LoongArch: KVM: Fix data race in kvm_ipi_regs_access cuitao
2026-04-14 1:50 ` Bibo Mao [this message]
2026-04-14 2:34 ` cuitao
2026-04-14 2:46 ` cuitao
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=ce3d170a-56a4-9343-49fa-2d8b3cbc9ebe@loongson.cn \
--to=maobibo@loongson.cn \
--cc=chenhuacai@kernel.org \
--cc=cuitao@kylinos.cn \
--cc=kernel@xen0n.name \
--cc=loongarch@lists.linux.dev \
--cc=zhaotianrui@loongson.cn \
/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;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).