linux-kernel - Re: [PATCH] xen: fix booting ballooned down hvm 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] [thread-next>] [day] [month] [year] [list]

Date:   Tue, 24 Oct 2017 10:33:07 -0400
From:   Boris Ostrovsky <boris.ostrovsky@...cle.com>
To:     Juergen Gross <jgross@...e.com>, linux-kernel@...r.kernel.org,
        xen-devel@...ts.xenproject.org
Cc:     hw42@...umj.de, stable@...r.kernel.org
Subject: Re: [PATCH] xen: fix booting ballooned down hvm guest

On 10/24/2017 04:10 AM, Juergen Gross wrote:
> Commit 96edd61dcf44362d3ef0bed1a5361e0ac7886a63 ("xen/balloon: don't
> online new memory initially") introduced a regression when booting a
> HVM domain with memory less than mem-max: instead of ballooning down
> immediately the system would try to use the memory up to mem-max
> resulting in Xen crashing the domain.
>
> For HVM domains the current size will be reflected in Xenstore node
> memory/static-max instead of memory/target.
>
> Additionally we have to trigger the ballooning process at once.
>
> Cc: <stable@...r.kernel.org> # 4.13
> Fixes: 96edd61dcf44362d3ef0bed1a5361e0ac7886a63 ("xen/balloon: don't
>        online new memory initially")
>
> Suggested-by: Boris Ostrovsky <boris.ostrovsky@...cle.com>
> Signed-off-by: Juergen Gross <jgross@...e.com>

Reported-by: HW42 <hw42@...umj.de>

> ---
>  drivers/xen/xen-balloon.c | 18 ++++++++++++------
>  1 file changed, 12 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/xen/xen-balloon.c b/drivers/xen/xen-balloon.c
> index e89136ab851e..3745748d9644 100644
> --- a/drivers/xen/xen-balloon.c
> +++ b/drivers/xen/xen-balloon.c
> @@ -57,7 +57,7 @@ static int register_balloon(struct device *dev);
>  static void watch_target(struct xenbus_watch *watch,
>  			 const char *path, const char *token)
>  {
> -	unsigned long long new_target;
> +	unsigned long long new_target, static_max;
>  	int err;
>  	static bool watch_fired;
>  	static long target_diff;
> @@ -72,13 +72,19 @@ static void watch_target(struct xenbus_watch *watch,
>  	 * pages. PAGE_SHIFT converts bytes to pages, hence PAGE_SHIFT - 10.
>  	 */
>  	new_target >>= PAGE_SHIFT - 10;
> -	if (watch_fired) {
> -		balloon_set_new_target(new_target - target_diff);
> -		return;
> +
> +	if (!watch_fired) {
> +		watch_fired = true;
> +		err = xenbus_scanf(XBT_NIL, "memory", "static-max", "%llu",
> +				   &static_max);
> +		if (err != 1)
> +			static_max = new_target;
> +		static_max >>= PAGE_SHIFT - 10;

if you set static_max to new_target you've already done the shift,
haven't you?


> +		target_diff = xen_pv_domain() ? 0

Why do we special-case PV?

-boris

> +				: static_max - balloon_stats.target_pages;
>  	}
>  
> -	watch_fired = true;
> -	target_diff = new_target - balloon_stats.target_pages;
> +	balloon_set_new_target(new_target - target_diff);
>  }
>  static struct xenbus_watch target_watch = {
>  	.node = "memory/target",