[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20220119223246.GA1539@anparri>
Date: Wed, 19 Jan 2022 23:32:46 +0100
From: Andrea Parri <parri.andrea@...il.com>
To: Yanming Liu <yanminglr@...il.com>
Cc: linux-hyperv@...r.kernel.org, linux-kernel@...r.kernel.org,
Andres Beltran <lkmlabelt@...il.com>,
Dexuan Cui <decui@...rosoft.com>, Wei Liu <wei.liu@...nel.org>,
Stephen Hemminger <sthemmin@...rosoft.com>,
Haiyang Zhang <haiyangz@...rosoft.com>,
"K. Y. Srinivasan" <kys@...rosoft.com>,
Michael Kelley <mikelley@...rosoft.com>
Subject: Re: [PATCH] Drivers: hv: balloon: account for vmbus packet header in
max_pkt_size
On Thu, Jan 20, 2022 at 04:20:52AM +0800, Yanming Liu wrote:
> Commit adae1e931acd ("Drivers: hv: vmbus: Copy packets sent by Hyper-V
> out of the ring buffer") introduced a notion of maximum packet size in
> vmbus channel and used that size to initialize a buffer holding all
> incoming packet along with their vmbus packet header. hv_balloon uses
> the default maximum packet size VMBUS_DEFAULT_MAX_PKT_SIZE which matches
> its maximum message size, however vmbus_open expects this size to also
> include vmbus packet header. This leads to 4096 bytes
> dm_unballoon_request messages being truncated to 4080 bytes. When the
> driver tries to read next packet it starts from a wrong read_index,
> receives garbage and prints a lot of "Unhandled message: type:
> <garbage>" in dmesg.
>
> Allocate the buffer with HV_HYP_PAGE_SIZE more bytes to make room for
> the header.
>
> Fixes: adae1e931acd ("Drivers: hv: vmbus: Copy packets sent by Hyper-V out of the ring buffer")
> Suggested-by: Michael Kelley (LINUX) <mikelley@...rosoft.com>
> Suggested-by: Andrea Parri (Microsoft) <parri.andrea@...il.com>
> Signed-off-by: Yanming Liu <yanminglr@...il.com>
> ---
> The patch was "[PATCH v2] hv: account for packet descriptor in maximum
> packet size". As pointed out by Michael Kelley [1], other hv drivers
> already overallocate a lot, and hv_balloon is hopefully the only
> remaining affected driver. It's better to just fix hv_balloon. Patch
> summary is changed to reflect this new (much smaller) scope.
hopefully - adeguately describing what/how we "know" (here), it remarks
us that our estimates are empirical (and that people may have different
opinions about "safe" estimates ;-))
Reviewed-by: Andrea Parri (Microsoft) <parri.andrea@...il.com>
Thanks,
Andrea
> [1] https://lore.kernel.org/linux-hyperv/CY4PR21MB1586D30C6CEC81EFC37A9848D7599@CY4PR21MB1586.namprd21.prod.outlook.com/
>
> drivers/hv/hv_balloon.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/drivers/hv/hv_balloon.c b/drivers/hv/hv_balloon.c
> index ca873a3b98db..f2d05bff4245 100644
> --- a/drivers/hv/hv_balloon.c
> +++ b/drivers/hv/hv_balloon.c
> @@ -1660,6 +1660,13 @@ static int balloon_connect_vsp(struct hv_device *dev)
> unsigned long t;
> int ret;
>
> + /*
> + * max_pkt_size should be large enough for one vmbus packet header plus
> + * our receive buffer size. Hyper-V sends messages up to
> + * HV_HYP_PAGE_SIZE bytes long on balloon channel.
> + */
> + dev->channel->max_pkt_size = HV_HYP_PAGE_SIZE * 2;
> +
> ret = vmbus_open(dev->channel, dm_ring_size, dm_ring_size, NULL, 0,
> balloon_onchannelcallback, dev);
> if (ret)
> --
> 2.34.1
>
Powered by blists - more mailing lists