Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Jeff Moyer <jmoyer@redhat.com>
To:     Seth Jenkins <sethjenkins@google.com>
Cc:     Alexander Viro <viro@zeniv.linux.org.uk>,
        Benjamin LaHaise <bcrl@kvack.org>,
        linux-fsdevel@vger.kernel.org, linux-aio@kvack.org,
        linux-kernel@vger.kernel.org, Jann Horn <jannh@google.com>,
        Pavel Emelyanov <xemul@parallels.com>, stable@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH] aio: fix mremap after fork null-deref
References: <20221104212519.538108-1-sethjenkins@google.com>
Date:   Wed, 11 Jan 2023 14:40:51 -0500
In-Reply-To: <20221104212519.538108-1-sethjenkins@google.com> (Seth Jenkins's
        message of "Fri, 4 Nov 2022 17:25:19 -0400")
Message-ID: <x49tu0wlv0c.fsf@segfault.boston.devel.redhat.com>
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Hi, Seth,

Seth Jenkins <sethjenkins@google.com> writes:

> Commit e4a0d3e720e7 ("aio: Make it possible to remap aio ring") introduced
> a null-deref if mremap is called on an old aio mapping after fork as
> mm->ioctx_table will be set to NULL.
>
> Fixes: e4a0d3e720e7 ("aio: Make it possible to remap aio ring")
> Cc: stable@vger.kernel.org
> Signed-off-by: Seth Jenkins <sethjenkins@google.com>
> ---
>  fs/aio.c | 20 +++++++++++---------
>  1 file changed, 11 insertions(+), 9 deletions(-)
>
> diff --git a/fs/aio.c b/fs/aio.c
> index 5b2ff20ad322..74eae7de7323 100644
> --- a/fs/aio.c
> +++ b/fs/aio.c
> @@ -361,16 +361,18 @@ static int aio_ring_mremap(struct vm_area_struct *vma)
>  	spin_lock(&mm->ioctx_lock);
>  	rcu_read_lock();
>  	table = rcu_dereference(mm->ioctx_table);
> -	for (i = 0; i < table->nr; i++) {
> -		struct kioctx *ctx;
> -
> -		ctx = rcu_dereference(table->table[i]);
> -		if (ctx && ctx->aio_ring_file == file) {
> -			if (!atomic_read(&ctx->dead)) {
> -				ctx->user_id = ctx->mmap_base = vma->vm_start;
> -				res = 0;
> +	if (table) {
> +		for (i = 0; i < table->nr; i++) {
> +			struct kioctx *ctx;
> +
> +			ctx = rcu_dereference(table->table[i]);
> +			if (ctx && ctx->aio_ring_file == file) {
> +				if (!atomic_read(&ctx->dead)) {
> +					ctx->user_id = ctx->mmap_base = vma->vm_start;
> +					res = 0;
> +				}
> +				break;
>  			}
> -			break;
>  		}
>  	}

I wonder if it would be better to not copy the ring mapping on fork.
Something like the below?  I think that would be more in line with
expectations (the ring isn't available in the child process).

-Jeff

diff --git a/fs/aio.c b/fs/aio.c
index 562916d85cba..dbf3b0749cb4 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -390,7 +390,7 @@ static const struct vm_operations_struct aio_ring_vm_ops = {
 
 static int aio_ring_mmap(struct file *file, struct vm_area_struct *vma)
 {
-	vma->vm_flags |= VM_DONTEXPAND;
+	vma->vm_flags |= VM_DONTEXPAND|VM_DONTCOPY;
 	vma->vm_ops = &aio_ring_vm_ops;
 	return 0;
 }