qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Jason Wang <jasowang@redhat.com>
To: Li Qiang <liq3ea@163.com>,
	dmitry.fleytman@gmail.com, pbonzini@redhat.com, mst@redhat.com
Cc: liq3ea@gmail.com, qemu-devel@nongnu.org
Subject: Re: [PATCH] e1000e: using bottom half to send packets
Date: Fri, 17 Jul 2020 11:10:28 +0800	[thread overview]
Message-ID: <281e3c85-b8eb-0c3e-afc3-41011861b8ea@redhat.com> (raw)
In-Reply-To: <20200716161453.61295-1-liq3ea@163.com>


On 2020/7/17 上午12:14, Li Qiang wrote:
> Alexander Bulekov reported a UAF bug related e1000e packets send.
>
> -->https://bugs.launchpad.net/qemu/+bug/1886362
>
> This is because the guest trigger a e1000e packet send and set the
> data's address to e1000e's MMIO address. So when the e1000e do DMA
> it will write the MMIO again and trigger re-entrancy and finally
> causes this UAF.
>
> Paolo suggested to use a bottom half whenever MMIO is doing complicate
> things in here:
> -->https://lists.nongnu.org/archive/html/qemu-devel/2020-07/msg03342.html
>
> Reference here:
> 'The easiest solution is to delay processing of descriptors to a bottom
> half whenever MMIO is doing something complicated.  This is also better
> for latency because it will free the vCPU thread more quickly and leave
> the work to the I/O thread.'


I think several things were missed in this patch (take virtio-net as a 
reference), do we need the following things:

- Cancel the bh when VM is stopped.
- A throttle to prevent bh from executing too much timer?
- A flag to record whether or not this a pending tx (and migrate it?)

Thanks


>
> This patch fixes this UAF.
>
> Signed-off-by: Li Qiang <liq3ea@163.com>
> ---
>   hw/net/e1000e_core.c | 25 +++++++++++++++++--------
>   hw/net/e1000e_core.h |  2 ++
>   2 files changed, 19 insertions(+), 8 deletions(-)
>
> diff --git a/hw/net/e1000e_core.c b/hw/net/e1000e_core.c
> index bcd186cac5..6165b04b68 100644
> --- a/hw/net/e1000e_core.c
> +++ b/hw/net/e1000e_core.c
> @@ -2423,32 +2423,27 @@ e1000e_set_dbal(E1000ECore *core, int index, uint32_t val)
>   static void
>   e1000e_set_tctl(E1000ECore *core, int index, uint32_t val)
>   {
> -    E1000E_TxRing txr;
>       core->mac[index] = val;
>   
>       if (core->mac[TARC0] & E1000_TARC_ENABLE) {
> -        e1000e_tx_ring_init(core, &txr, 0);
> -        e1000e_start_xmit(core, &txr);
> +        qemu_bh_schedule(core->tx[0].tx_bh);
>       }
>   
>       if (core->mac[TARC1] & E1000_TARC_ENABLE) {
> -        e1000e_tx_ring_init(core, &txr, 1);
> -        e1000e_start_xmit(core, &txr);
> +        qemu_bh_schedule(core->tx[1].tx_bh);
>       }
>   }
>   
>   static void
>   e1000e_set_tdt(E1000ECore *core, int index, uint32_t val)
>   {
> -    E1000E_TxRing txr;
>       int qidx = e1000e_mq_queue_idx(TDT, index);
>       uint32_t tarc_reg = (qidx == 0) ? TARC0 : TARC1;
>   
>       core->mac[index] = val & 0xffff;
>   
>       if (core->mac[tarc_reg] & E1000_TARC_ENABLE) {
> -        e1000e_tx_ring_init(core, &txr, qidx);
> -        e1000e_start_xmit(core, &txr);
> +        qemu_bh_schedule(core->tx[qidx].tx_bh);
>       }
>   }
>   
> @@ -3322,6 +3317,16 @@ e1000e_vm_state_change(void *opaque, int running, RunState state)
>       }
>   }
>   
> +static void e1000e_core_tx_bh(void *opaque)
> +{
> +    struct e1000e_tx *tx = opaque;
> +    E1000ECore *core = tx->core;
> +    E1000E_TxRing txr;
> +
> +    e1000e_tx_ring_init(core, &txr, tx - &core->tx[0]);
> +    e1000e_start_xmit(core, &txr);
> +}
> +
>   void
>   e1000e_core_pci_realize(E1000ECore     *core,
>                           const uint16_t *eeprom_templ,
> @@ -3340,6 +3345,8 @@ e1000e_core_pci_realize(E1000ECore     *core,
>       for (i = 0; i < E1000E_NUM_QUEUES; i++) {
>           net_tx_pkt_init(&core->tx[i].tx_pkt, core->owner,
>                           E1000E_MAX_TX_FRAGS, core->has_vnet);
> +        core->tx[i].core = core;
> +        core->tx[i].tx_bh = qemu_bh_new(e1000e_core_tx_bh, &core->tx[i]);
>       }
>   
>       net_rx_pkt_init(&core->rx_pkt, core->has_vnet);
> @@ -3367,6 +3374,8 @@ e1000e_core_pci_uninit(E1000ECore *core)
>       for (i = 0; i < E1000E_NUM_QUEUES; i++) {
>           net_tx_pkt_reset(core->tx[i].tx_pkt);
>           net_tx_pkt_uninit(core->tx[i].tx_pkt);
> +        qemu_bh_delete(core->tx[i].tx_bh);
> +        core->tx[i].tx_bh = NULL;
>       }
>   
>       net_rx_pkt_uninit(core->rx_pkt);
> diff --git a/hw/net/e1000e_core.h b/hw/net/e1000e_core.h
> index aee32f7e48..94ddc6afc2 100644
> --- a/hw/net/e1000e_core.h
> +++ b/hw/net/e1000e_core.h
> @@ -77,6 +77,8 @@ struct E1000Core {
>           unsigned char sum_needed;
>           bool cptse;
>           struct NetTxPkt *tx_pkt;
> +        QEMUBH *tx_bh;
> +        E1000ECore *core;
>       } tx[E1000E_NUM_QUEUES];
>   
>       struct NetRxPkt *rx_pkt;



  reply	other threads:[~2020-07-17  3:11 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-16 16:14 [PATCH] e1000e: using bottom half to send packets Li Qiang
2020-07-17  3:10 ` Jason Wang [this message]
2020-07-17  4:46   ` Li Qiang
2020-07-17  5:38     ` Jason Wang
2020-07-17 15:46       ` Li Qiang
2020-07-20  3:59         ` Jason Wang
2020-07-20  4:41           ` Li Qiang
2020-07-17 15:52   ` Peter Maydell
2020-07-20  4:00     ` 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=281e3c85-b8eb-0c3e-afc3-41011861b8ea@redhat.com \
    --to=jasowang@redhat.com \
    --cc=dmitry.fleytman@gmail.com \
    --cc=liq3ea@163.com \
    --cc=liq3ea@gmail.com \
    --cc=mst@redhat.com \
    --cc=pbonzini@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 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 NNTP newsgroup(s).