All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dmitry Fleytman <dmitry@daynix.com>
To: P J P <ppandit@redhat.com>
Cc: Qinghao Tang <luodalongde@gmail.com>,
	Jason Wang <jasowang@redhat.com>,
	qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] net: vmxnet3: memory leakage issue
Date: Sun, 13 Dec 2015 11:45:24 +0200	[thread overview]
Message-ID: <1C7745A5-54CD-439B-8A4B-442D97C58823@daynix.com> (raw)
In-Reply-To: <alpine.LFD.2.20.1512091713000.22387@wniryva>

[-- Attachment #1: Type: text/plain, Size: 7206 bytes --]

Hello Prasad,

On 9 Dec 2015, at 17:28 PM, P J P <ppandit@redhat.com> wrote:
> 
>   Hello Jason, Dmitry,
> 
> +-- On Tue, 8 Dec 2015, P J P wrote --+
> | |1) VMXNET3_CMD_QUIESCE_DEV
> | 
> |   IIUC, it is used to pause the device when the receiver end is unable to 
> | keee-up with the incoming flow. After a brief period, the operation could be 
> | resumed again.
> | 
> | |2) VMXNET3_REG_DSAL
> | 
> |    Shared memory between a driver and the device appears to be set in two 
> | steps. Firs low address, followed by the high address(VMXNET3_REG_DSAH). I 
> | guess 's->device_active' needs to be enabled again while setting the higher 
> | part of the address.
> 
> Please see below another (tested)patch, it fixes the VMXNET3_CMD_QUIESCE_DEV 
> case above. It's not clear if the device would be initialised and active while 
> setting shared memory via VMXNET3_REG_DSAL/DSAH.
> 
> ===
> From 81c4ecb67635435f01397dc21210497e6420bdca Mon Sep 17 00:00:00 2001
> From: Prasad J Pandit <pjp@fedoraproject.org>
> Date: Wed, 9 Dec 2015 20:45:25 +0530
> Subject: [PATCH] net: vmxnet3: avoid memory leakage in activate_device
> 
> Vmxnet3 device emulator does not check if the device is active
> before activating it, resulting in memory leakage on the host.
> This memory leakage could also occur, if the device was paused
> during flow control and activated thereafter.
> 
> Introduced a 'Pause' state for the vmxnet3 device to differentiate
> it from the earlier active and inactive states. One need not
> 'activate' the device to resume operation after pause.
> 
> This patch adds a check to verify these device states and avoid
> memory leakage during activating the device.
> 
> Reported-by: Qinghao Tang <luodalongde@gmail.com>
> Signed-off-by: Prasad J Pandit <pjp@fedoraproject.org>
> ---
> hw/net/vmxnet3.c | 43 +++++++++++++++++++++++++++++++++++--------
> hw/net/vmxnet3.h |  6 ++++++
> 2 files changed, 41 insertions(+), 8 deletions(-)
> 
> diff --git a/hw/net/vmxnet3.c b/hw/net/vmxnet3.c
> index 37373e5..be2c9e2 100644
> --- a/hw/net/vmxnet3.c
> +++ b/hw/net/vmxnet3.c
> @@ -1195,7 +1195,19 @@ static void vmxnet3_reset_mac(VMXNET3State *s)
> static void vmxnet3_deactivate_device(VMXNET3State *s)
> {
>     VMW_CBPRN("Deactivating vmxnet3...");
> -    s->device_active = false;
> +    s->device_active = VMXNET3_DEV_DEACTIVE;
> +}
> +
> +static void vmxnet3_pause_device(VMXNET3State *s)
> +{
> +    VMW_CBPRN("Pausing vmxnet3...");
> +    s->device_active = VMXNET3_DEV_PAUSE;
> +}
> +
> +static void vmxnet3_resume_device(VMXNET3State *s)
> +{
> +    VMW_CBPRN("Resuming vmxnet3...");
> +    s->device_active = VMXNET3_DEV_ACTIVE;
> }
> 
> static void vmxnet3_reset(VMXNET3State *s)
> @@ -1431,6 +1443,12 @@ static void vmxnet3_activate_device(VMXNET3State *s)
>         return;
>     }
> 
> +    /* Verify if device is active */
> +    if (s->device_active) {
> +        VMW_CFPRN("Vmxnet3 device is active");
> +        return;
> +    }
> +
>     vmxnet3_adjust_by_guest_type(s);
>     vmxnet3_update_features(s);
>     vmxnet3_update_pm_state(s);
> @@ -1566,7 +1584,7 @@ static void vmxnet3_activate_device(VMXNET3State *s)
> 
>     vmxnet3_reset_mac(s);
> 
> -    s->device_active = true;
> +    s->device_active = VMXNET3_DEV_ACTIVE;
> }
> 
> static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
> @@ -1627,8 +1645,13 @@ static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
>         break;
> 
>     case VMXNET3_CMD_QUIESCE_DEV:
> -        VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - pause the device");
> -        vmxnet3_deactivate_device(s);
> +        if (s->device_active & VMXNET3_DEV_ACTIVE) {
> +            VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - pause the device");
> +            vmxnet3_pause_device(s);
> +        } else if (s->device_active & VMXNET3_DEV_PAUSE) {
> +            VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - resume the device");
> +            vmxnet3_resume_device(s);
> +        }

According to Linux driver code VMXNET3_CMD_QUIESCE_DEV does not flip paused/active states.
It always disables device, see vmxnet3_resume() for example (https://lxr.missinglinkelectronics.com/linux/drivers/net/vmxnet3/vmxnet3_drv.c#L3423 <https://lxr.missinglinkelectronics.com/linux/drivers/net/vmxnet3/vmxnet3_drv.c#L3423>):

Driver issues VMXNET3_CMD_QUIESCE_DEV to clear the device state and then performs activate sequence to launch the device. 

So the correct fix should:

1. On device activation: check if device is active - do nothing
2. In all places that set device_active to false, i.e. device quiesce, reset and VMXNET3_REG_DSAL set to zero: deallocate tx/rx packets as done in vmxnet3_net_uninit():

net_tx_pkt_reset(s->tx_pkt);
net_tx_pkt_uninit(s->tx_pkt);
net_rx_pkt_uninit(s->rx_pkt);

It could be a good idea to extend vmxnet3_deactivate_device() with those lines and call it from every place that sets device_active to false or frees TX/RX packets.

Best Regards,
Dmitry

>         break;
> 
>     case VMXNET3_CMD_GET_CONF_INTR:
> @@ -1652,12 +1675,16 @@ static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
> 
>     switch (s->last_command) {
>     case VMXNET3_CMD_ACTIVATE_DEV:
> -        ret = (s->device_active) ? 0 : -1;
> +        ret = (s->device_active & VMXNET3_DEV_ACTIVE) ? 0 : -1;
>         VMW_CFPRN("Device active: %" PRIx64, ret);
>         break;
> 
> -    case VMXNET3_CMD_RESET_DEV:
>     case VMXNET3_CMD_QUIESCE_DEV:
> +        ret = (s->device_active & VMXNET3_DEV_PAUSE) ? 0 : -1;
> +        VMW_CFPRN("Device pause: %" PRIx64, ret);
> +        break;
> +
> +    case VMXNET3_CMD_RESET_DEV:
>     case VMXNET3_CMD_GET_QUEUE_STATUS:
>         ret = 0;
>         break;
> @@ -1741,7 +1768,7 @@ vmxnet3_io_bar1_write(void *opaque,
>          * shared address only after we get the high part
>          */
>         if (val == 0) {
> -            s->device_active = false;
> +            s->device_active = VMXNET3_DEV_DEACTIVE;
>         }
>         s->temp_shared_guest_driver_memory = val;
>         s->drv_shmem = 0;
> @@ -1863,7 +1890,7 @@ static int
> vmxnet3_can_receive(NetClientState *nc)
> {
>     VMXNET3State *s = qemu_get_nic_opaque(nc);
> -    return s->device_active &&
> +    return (s->device_active & VMXNET3_DEV_ACTIVE) &&
>            VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
> }
> 
> diff --git a/hw/net/vmxnet3.h b/hw/net/vmxnet3.h
> index f7006af..71a7861 100644
> --- a/hw/net/vmxnet3.h
> +++ b/hw/net/vmxnet3.h
> @@ -202,6 +202,12 @@ enum {
>     VMXNET3_CMD_GET_ADAPTIVE_RING_INFO                    /* 0xF00D0009 */
> };
> 
> +enum {
> +    VMXNET3_DEV_DEACTIVE = 0x00,  /* device deactive */
> +    VMXNET3_DEV_ACTIVE = 0x01,    /* device active  */
> +    VMXNET3_DEV_PAUSE = 0x02      /* device pause */
> +};
> +
> /* Adaptive Ring Info Flags */
> #define VMXNET3_DISABLE_ADAPTIVE_RING 1
> 
> -- 
> 2.4.3
> === 
> 
> 
> 
> Thank you.
> --
> Prasad J Pandit / Red Hat Product Security Team
> 47AF CE69 3A90 54AA 9045 1053 DD13 3D32 FE5B 041F


[-- Attachment #2: Type: text/html, Size: 12060 bytes --]

  parent reply	other threads:[~2015-12-13  9:45 UTC|newest]

Thread overview: 21+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-12-02 12:17 [Qemu-devel] net: vmxnet3: memory leakage issue P J P
2015-12-03  7:17 ` Dmitry Fleytman
2015-12-03 11:20   ` P J P
2015-12-04  3:16   ` Jason Wang
2015-12-08 10:17     ` P J P
2015-12-09 15:28       ` P J P
2015-12-11  9:10         ` Jason Wang
2015-12-11  9:34           ` Dmitry Fleytman
2015-12-11 10:04           ` P J P
2015-12-13  8:27             ` Dmitry Fleytman
2015-12-13  9:45         ` Dmitry Fleytman [this message]
2015-12-14 11:58           ` P J P
2015-12-14 17:27             ` Dmitry Fleytman
2015-12-15  6:57               ` P J P
2015-12-15  7:01                 ` Dmitry Fleytman
2015-12-15  8:00                   ` P J P
2015-12-15  8:24                     ` Jason Wang
2015-12-15  8:50                       ` P J P
2015-12-15  8:43             ` Miao Yan
2015-12-15 10:08               ` P J P
2015-12-04  2:22 ` Jason Wang

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=1C7745A5-54CD-439B-8A4B-442D97C58823@daynix.com \
    --to=dmitry@daynix.com \
    --cc=jasowang@redhat.com \
    --cc=luodalongde@gmail.com \
    --cc=ppandit@redhat.com \
    --cc=qemu-devel@nongnu.org \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.