In drivers/net/ethernet/nvidia/forcedeth.c, the functions
nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
executed with nv_poll_controller().
nv_start_xmit
line 2321: prev_tx_ctx->skb = skb;
nv_start_xmit_optimized
line 2479: prev_tx_ctx->skb = skb;
nv_poll_controller
nv_do_nic_poll
line 4134: spin_lock(&np->lock);
nv_drain_rxtx
nv_drain_tx
nv_release_txskb
line 2004: dev_kfree_skb_any(tx_skb->skb);
Thus, two possible concurrency use-after-free bugs may occur.
To fix these possible bugs, the calls to spin_lock_irqsave() in
nv_start_xmit() and nv_start_xmit_optimized() are moved to the
front of "prev_tx_ctx->skb = skb;"
Signed-off-by: Jia-Ju Bai <[email protected]>
---
drivers/net/ethernet/nvidia/forcedeth.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/drivers/net/ethernet/nvidia/forcedeth.c b/drivers/net/ethernet/nvidia/forcedeth.c
index 1d9b0d44ddb6..48fa5a0bd2cb 100644
--- a/drivers/net/ethernet/nvidia/forcedeth.c
+++ b/drivers/net/ethernet/nvidia/forcedeth.c
@@ -2317,6 +2317,8 @@ static netdev_tx_t nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
/* set last fragment flag */
prev_tx->flaglen |= cpu_to_le32(tx_flags_extra);
+ spin_lock_irqsave(&np->lock, flags);
+
/* save skb in this slot's context area */
prev_tx_ctx->skb = skb;
@@ -2326,8 +2328,6 @@ static netdev_tx_t nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
tx_flags_extra = skb->ip_summed == CHECKSUM_PARTIAL ?
NV_TX2_CHECKSUM_L3 | NV_TX2_CHECKSUM_L4 : 0;
- spin_lock_irqsave(&np->lock, flags);
-
/* set tx flags */
start_tx->flaglen |= cpu_to_le32(tx_flags | tx_flags_extra);
@@ -2475,6 +2475,8 @@ static netdev_tx_t nv_start_xmit_optimized(struct sk_buff *skb,
/* set last fragment flag */
prev_tx->flaglen |= cpu_to_le32(NV_TX2_LASTPACKET);
+ spin_lock_irqsave(&np->lock, flags);
+
/* save skb in this slot's context area */
prev_tx_ctx->skb = skb;
@@ -2491,8 +2493,6 @@ static netdev_tx_t nv_start_xmit_optimized(struct sk_buff *skb,
else
start_tx->txvlan = 0;
- spin_lock_irqsave(&np->lock, flags);
-
if (np->tx_limit) {
/* Limit the number of outstanding tx. Setup all fragments, but
* do not set the VALID bit on the first descriptor. Save a pointer
--
2.17.0
?? 2019/1/8 20:45, Jia-Ju Bai д??:
> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
> executed with nv_poll_controller().
>
> nv_start_xmit
> line 2321: prev_tx_ctx->skb = skb;
>
> nv_start_xmit_optimized
> line 2479: prev_tx_ctx->skb = skb;
>
> nv_poll_controller
> nv_do_nic_poll
> line 4134: spin_lock(&np->lock);
> nv_drain_rxtx
> nv_drain_tx
> nv_release_txskb
> line 2004: dev_kfree_skb_any(tx_skb->skb);
>
> Thus, two possible concurrency use-after-free bugs may occur.
>
> To fix these possible bugs,
Does this really occur? Can you reproduce this ?
> the calls to spin_lock_irqsave() in
> nv_start_xmit() and nv_start_xmit_optimized() are moved to the
> front of "prev_tx_ctx->skb = skb;"
>
> Signed-off-by: Jia-Ju Bai <[email protected]>
> ---
> drivers/net/ethernet/nvidia/forcedeth.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/ethernet/nvidia/forcedeth.c b/drivers/net/ethernet/nvidia/forcedeth.c
> index 1d9b0d44ddb6..48fa5a0bd2cb 100644
> --- a/drivers/net/ethernet/nvidia/forcedeth.c
> +++ b/drivers/net/ethernet/nvidia/forcedeth.c
> @@ -2317,6 +2317,8 @@ static netdev_tx_t nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
> /* set last fragment flag */
> prev_tx->flaglen |= cpu_to_le32(tx_flags_extra);
>
> + spin_lock_irqsave(&np->lock, flags);
> +
> /* save skb in this slot's context area */
> prev_tx_ctx->skb = skb;
>
> @@ -2326,8 +2328,6 @@ static netdev_tx_t nv_start_xmit(struct sk_buff *skb, struct net_device *dev)
> tx_flags_extra = skb->ip_summed == CHECKSUM_PARTIAL ?
> NV_TX2_CHECKSUM_L3 | NV_TX2_CHECKSUM_L4 : 0;
>
> - spin_lock_irqsave(&np->lock, flags);
> -
> /* set tx flags */
> start_tx->flaglen |= cpu_to_le32(tx_flags | tx_flags_extra);
>
> @@ -2475,6 +2475,8 @@ static netdev_tx_t nv_start_xmit_optimized(struct sk_buff *skb,
> /* set last fragment flag */
> prev_tx->flaglen |= cpu_to_le32(NV_TX2_LASTPACKET);
>
> + spin_lock_irqsave(&np->lock, flags);
> +
> /* save skb in this slot's context area */
> prev_tx_ctx->skb = skb;
>
> @@ -2491,8 +2493,6 @@ static netdev_tx_t nv_start_xmit_optimized(struct sk_buff *skb,
> else
> start_tx->txvlan = 0;
>
> - spin_lock_irqsave(&np->lock, flags);
> -
> if (np->tx_limit) {
> /* Limit the number of outstanding tx. Setup all fragments, but
> * do not set the VALID bit on the first descriptor. Save a pointer
On 2019/1/8 20:54, Zhu Yanjun wrote:
>
> 在 2019/1/8 20:45, Jia-Ju Bai 写道:
>> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
>> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
>> executed with nv_poll_controller().
>>
>> nv_start_xmit
>> line 2321: prev_tx_ctx->skb = skb;
>>
>> nv_start_xmit_optimized
>> line 2479: prev_tx_ctx->skb = skb;
>>
>> nv_poll_controller
>> nv_do_nic_poll
>> line 4134: spin_lock(&np->lock);
>> nv_drain_rxtx
>> nv_drain_tx
>> nv_release_txskb
>> line 2004: dev_kfree_skb_any(tx_skb->skb);
>>
>> Thus, two possible concurrency use-after-free bugs may occur.
>>
>> To fix these possible bugs,
>
>
> Does this really occur? Can you reproduce this ?
This bug is not found by the real execution.
It is found by a static tool written by myself, and then I check it by
manual code review.
Best wishes,
Jia-Ju Bai
On 2019/1/8 20:57, Jia-Ju Bai wrote:
>
>
> On 2019/1/8 20:54, Zhu Yanjun wrote:
>>
>> 在 2019/1/8 20:45, Jia-Ju Bai 写道:
>>> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
>>> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
>>> executed with nv_poll_controller().
>>>
>>> nv_start_xmit
>>> line 2321: prev_tx_ctx->skb = skb;
>>>
>>> nv_start_xmit_optimized
>>> line 2479: prev_tx_ctx->skb = skb;
>>>
>>> nv_poll_controller
>>> nv_do_nic_poll
>>> line 4134: spin_lock(&np->lock);
>>> nv_drain_rxtx
>>> nv_drain_tx
>>> nv_release_txskb
>>> line 2004: dev_kfree_skb_any(tx_skb->skb);
>>>
>>> Thus, two possible concurrency use-after-free bugs may occur.
>>>
>>> To fix these possible bugs,
>>
>>
>> Does this really occur? Can you reproduce this ?
>
> This bug is not found by the real execution.
> It is found by a static tool written by myself, and then I check it by
> manual code review.
Before "line 2004: dev_kfree_skb_any(tx_skb->skb); ",
"
nv_disable_irq(dev);
nv_napi_disable(dev);
netif_tx_lock_bh(dev);
netif_addr_lock(dev);
spin_lock(&np->lock);
/* stop engines */
nv_stop_rxtx(dev); <---this stop rxtx
nv_txrx_reset(dev);
"
In this case, does nv_start_xmit or nv_start_xmit_optimized still work well?
Zhu Yanjun
>
>
> Best wishes,
> Jia-Ju Bai
On 2019/1/9 9:24, Yanjun Zhu wrote:
>
> On 2019/1/8 20:57, Jia-Ju Bai wrote:
>>
>>
>> On 2019/1/8 20:54, Zhu Yanjun wrote:
>>>
>>> 在 2019/1/8 20:45, Jia-Ju Bai 写道:
>>>> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
>>>> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
>>>> executed with nv_poll_controller().
>>>>
>>>> nv_start_xmit
>>>> line 2321: prev_tx_ctx->skb = skb;
>>>>
>>>> nv_start_xmit_optimized
>>>> line 2479: prev_tx_ctx->skb = skb;
>>>>
>>>> nv_poll_controller
>>>> nv_do_nic_poll
>>>> line 4134: spin_lock(&np->lock);
>>>> nv_drain_rxtx
>>>> nv_drain_tx
>>>> nv_release_txskb
>>>> line 2004: dev_kfree_skb_any(tx_skb->skb);
>>>>
>>>> Thus, two possible concurrency use-after-free bugs may occur.
>>>>
>>>> To fix these possible bugs,
>>>
>>>
>>> Does this really occur? Can you reproduce this ?
>>
>> This bug is not found by the real execution.
>> It is found by a static tool written by myself, and then I check it
>> by manual code review.
>
> Before "line 2004: dev_kfree_skb_any(tx_skb->skb); ",
>
> "
>
> nv_disable_irq(dev);
> nv_napi_disable(dev);
> netif_tx_lock_bh(dev);
> netif_addr_lock(dev);
> spin_lock(&np->lock);
> /* stop engines */
> nv_stop_rxtx(dev); <---this stop rxtx
> nv_txrx_reset(dev);
> "
>
> In this case, does nv_start_xmit or nv_start_xmit_optimized still work
> well?
>
nv_stop_rxtx() calls nv_stop_tx(dev).
static void nv_stop_tx(struct net_device *dev)
{
struct fe_priv *np = netdev_priv(dev);
u8 __iomem *base = get_hwbase(dev);
u32 tx_ctrl = readl(base + NvRegTransmitterControl);
if (!np->mac_in_use)
tx_ctrl &= ~NVREG_XMITCTL_START;
else
tx_ctrl |= NVREG_XMITCTL_TX_PATH_EN;
writel(tx_ctrl, base + NvRegTransmitterControl);
if (reg_delay(dev, NvRegTransmitterStatus, NVREG_XMITSTAT_BUSY, 0,
NV_TXSTOP_DELAY1, NV_TXSTOP_DELAY1MAX))
netdev_info(dev, "%s: TransmitterStatus remained busy\n",
__func__);
udelay(NV_TXSTOP_DELAY2);
if (!np->mac_in_use)
writel(readl(base + NvRegTransmitPoll) &
NVREG_TRANSMITPOLL_MAC_ADDR_REV,
base + NvRegTransmitPoll);
}
nv_stop_tx() seems to only write registers to stop transmitting for
hardware.
But it does not wait until nv_start_xmit() and nv_start_xmit_optimized()
finish execution.
Maybe netif_stop_queue() should be used here to stop transmitting for
network layer, but this function does not seem to wait, either.
Do you know any function that can wait until ".ndo_start_xmit" finish
execution?
Best wishes,
Jia-Ju Bai
On 2019/1/9 10:03, Jia-Ju Bai wrote:
>
>
> On 2019/1/9 9:24, Yanjun Zhu wrote:
>>
>> On 2019/1/8 20:57, Jia-Ju Bai wrote:
>>>
>>>
>>> On 2019/1/8 20:54, Zhu Yanjun wrote:
>>>>
>>>> 在 2019/1/8 20:45, Jia-Ju Bai 写道:
>>>>> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
>>>>> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
>>>>> executed with nv_poll_controller().
>>>>>
>>>>> nv_start_xmit
>>>>> line 2321: prev_tx_ctx->skb = skb;
>>>>>
>>>>> nv_start_xmit_optimized
>>>>> line 2479: prev_tx_ctx->skb = skb;
>>>>>
>>>>> nv_poll_controller
>>>>> nv_do_nic_poll
>>>>> line 4134: spin_lock(&np->lock);
>>>>> nv_drain_rxtx
>>>>> nv_drain_tx
>>>>> nv_release_txskb
>>>>> line 2004: dev_kfree_skb_any(tx_skb->skb);
>>>>>
>>>>> Thus, two possible concurrency use-after-free bugs may occur.
>>>>>
>>>>> To fix these possible bugs,
>>>>
>>>>
>>>> Does this really occur? Can you reproduce this ?
>>>
>>> This bug is not found by the real execution.
>>> It is found by a static tool written by myself, and then I check it
>>> by manual code review.
>>
>> Before "line 2004: dev_kfree_skb_any(tx_skb->skb); ",
>>
>> "
>>
>> nv_disable_irq(dev);
>> nv_napi_disable(dev);
>> netif_tx_lock_bh(dev);
>> netif_addr_lock(dev);
>> spin_lock(&np->lock);
>> /* stop engines */
>> nv_stop_rxtx(dev); <---this stop rxtx
>> nv_txrx_reset(dev);
>> "
>>
>> In this case, does nv_start_xmit or nv_start_xmit_optimized still
>> work well?
>>
>
> nv_stop_rxtx() calls nv_stop_tx(dev).
>
> static void nv_stop_tx(struct net_device *dev)
> {
> struct fe_priv *np = netdev_priv(dev);
> u8 __iomem *base = get_hwbase(dev);
> u32 tx_ctrl = readl(base + NvRegTransmitterControl);
>
> if (!np->mac_in_use)
> tx_ctrl &= ~NVREG_XMITCTL_START;
> else
> tx_ctrl |= NVREG_XMITCTL_TX_PATH_EN;
> writel(tx_ctrl, base + NvRegTransmitterControl);
> if (reg_delay(dev, NvRegTransmitterStatus, NVREG_XMITSTAT_BUSY, 0,
> NV_TXSTOP_DELAY1, NV_TXSTOP_DELAY1MAX))
> netdev_info(dev, "%s: TransmitterStatus remained busy\n",
> __func__);
>
> udelay(NV_TXSTOP_DELAY2);
> if (!np->mac_in_use)
> writel(readl(base + NvRegTransmitPoll) &
> NVREG_TRANSMITPOLL_MAC_ADDR_REV,
> base + NvRegTransmitPoll);
> }
>
> nv_stop_tx() seems to only write registers to stop transmitting for
> hardware.
> But it does not wait until nv_start_xmit() and
> nv_start_xmit_optimized() finish execution.
There are 3 modes in forcedeth NIC.
In throughput mode (0), every tx & rx packet will generate an interrupt.
In CPU mode (1), interrupts are controlled by a timer.
In dynamic mode (2), the mode toggles between throughput and CPU mode
based on network load.
From the source code,
"np->recover_error = 1;" is related with CPU mode.
nv_start_xmit or nv_start_xmit_optimized seems related with ghroughput mode.
In static void nv_do_nic_poll(struct timer_list *t),
when if (np->recover_error), line 2004: dev_kfree_skb_any(tx_skb->skb);
will run.
When "np->recover_error=1", do you think nv_start_xmit or
nv_start_xmit_optimized will be called?
> Maybe netif_stop_queue() should be used here to stop transmitting for
> network layer, but this function does not seem to wait, either.
> Do you know any function that can wait until ".ndo_start_xmit" finish
> execution?
>
>
> Best wishes,
> Jia-Ju Bai
>
>
On 2019/1/9 10:35, Yanjun Zhu wrote:
>
> On 2019/1/9 10:03, Jia-Ju Bai wrote:
>>
>>
>> On 2019/1/9 9:24, Yanjun Zhu wrote:
>>>
>>> On 2019/1/8 20:57, Jia-Ju Bai wrote:
>>>>
>>>>
>>>> On 2019/1/8 20:54, Zhu Yanjun wrote:
>>>>>
>>>>> 在 2019/1/8 20:45, Jia-Ju Bai 写道:
>>>>>> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
>>>>>> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
>>>>>> executed with nv_poll_controller().
>>>>>>
>>>>>> nv_start_xmit
>>>>>> line 2321: prev_tx_ctx->skb = skb;
>>>>>>
>>>>>> nv_start_xmit_optimized
>>>>>> line 2479: prev_tx_ctx->skb = skb;
>>>>>>
>>>>>> nv_poll_controller
>>>>>> nv_do_nic_poll
>>>>>> line 4134: spin_lock(&np->lock);
>>>>>> nv_drain_rxtx
>>>>>> nv_drain_tx
>>>>>> nv_release_txskb
>>>>>> line 2004: dev_kfree_skb_any(tx_skb->skb);
>>>>>>
>>>>>> Thus, two possible concurrency use-after-free bugs may occur.
>>>>>>
>>>>>> To fix these possible bugs,
>>>>>
>>>>>
>>>>> Does this really occur? Can you reproduce this ?
>>>>
>>>> This bug is not found by the real execution.
>>>> It is found by a static tool written by myself, and then I check it
>>>> by manual code review.
>>>
>>> Before "line 2004: dev_kfree_skb_any(tx_skb->skb); ",
>>>
>>> "
>>>
>>> nv_disable_irq(dev);
>>> nv_napi_disable(dev);
>>> netif_tx_lock_bh(dev);
>>> netif_addr_lock(dev);
>>> spin_lock(&np->lock);
>>> /* stop engines */
>>> nv_stop_rxtx(dev); <---this stop rxtx
>>> nv_txrx_reset(dev);
>>> "
>>>
>>> In this case, does nv_start_xmit or nv_start_xmit_optimized still
>>> work well?
>>>
>>
>> nv_stop_rxtx() calls nv_stop_tx(dev).
>>
>> static void nv_stop_tx(struct net_device *dev)
>> {
>> struct fe_priv *np = netdev_priv(dev);
>> u8 __iomem *base = get_hwbase(dev);
>> u32 tx_ctrl = readl(base + NvRegTransmitterControl);
>>
>> if (!np->mac_in_use)
>> tx_ctrl &= ~NVREG_XMITCTL_START;
>> else
>> tx_ctrl |= NVREG_XMITCTL_TX_PATH_EN;
>> writel(tx_ctrl, base + NvRegTransmitterControl);
>> if (reg_delay(dev, NvRegTransmitterStatus, NVREG_XMITSTAT_BUSY, 0,
>> NV_TXSTOP_DELAY1, NV_TXSTOP_DELAY1MAX))
>> netdev_info(dev, "%s: TransmitterStatus remained busy\n",
>> __func__);
>>
>> udelay(NV_TXSTOP_DELAY2);
>> if (!np->mac_in_use)
>> writel(readl(base + NvRegTransmitPoll) &
>> NVREG_TRANSMITPOLL_MAC_ADDR_REV,
>> base + NvRegTransmitPoll);
>> }
>>
>> nv_stop_tx() seems to only write registers to stop transmitting for
>> hardware.
>> But it does not wait until nv_start_xmit() and
>> nv_start_xmit_optimized() finish execution.
> There are 3 modes in forcedeth NIC.
> In throughput mode (0), every tx & rx packet will generate an interrupt.
> In CPU mode (1), interrupts are controlled by a timer.
> In dynamic mode (2), the mode toggles between throughput and CPU mode
> based on network load.
>
> From the source code,
>
> "np->recover_error = 1;" is related with CPU mode.
>
> nv_start_xmit or nv_start_xmit_optimized seems related with ghroughput
> mode.
>
> In static void nv_do_nic_poll(struct timer_list *t),
> when if (np->recover_error), line 2004:
> dev_kfree_skb_any(tx_skb->skb); will run.
>
> When "np->recover_error=1", do you think nv_start_xmit or
> nv_start_xmit_optimized will be called?
Sorry, I do not know about these modes...
But I still think nv_start_xmit() or nv_start_xmit_optimized() can be
called here, in no matter which mode :)
Best wishes,
Jia-Ju Bai
On 2019/1/9 11:20, Jia-Ju Bai wrote:
>
>
> On 2019/1/9 10:35, Yanjun Zhu wrote:
>>
>> On 2019/1/9 10:03, Jia-Ju Bai wrote:
>>>
>>>
>>> On 2019/1/9 9:24, Yanjun Zhu wrote:
>>>>
>>>> On 2019/1/8 20:57, Jia-Ju Bai wrote:
>>>>>
>>>>>
>>>>> On 2019/1/8 20:54, Zhu Yanjun wrote:
>>>>>>
>>>>>> 在 2019/1/8 20:45, Jia-Ju Bai 写道:
>>>>>>> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
>>>>>>> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
>>>>>>> executed with nv_poll_controller().
>>>>>>>
>>>>>>> nv_start_xmit
>>>>>>> line 2321: prev_tx_ctx->skb = skb;
>>>>>>>
>>>>>>> nv_start_xmit_optimized
>>>>>>> line 2479: prev_tx_ctx->skb = skb;
>>>>>>>
>>>>>>> nv_poll_controller
>>>>>>> nv_do_nic_poll
>>>>>>> line 4134: spin_lock(&np->lock);
>>>>>>> nv_drain_rxtx
>>>>>>> nv_drain_tx
>>>>>>> nv_release_txskb
>>>>>>> line 2004: dev_kfree_skb_any(tx_skb->skb);
>>>>>>>
>>>>>>> Thus, two possible concurrency use-after-free bugs may occur.
>>>>>>>
>>>>>>> To fix these possible bugs,
>>>>>>
>>>>>>
>>>>>> Does this really occur? Can you reproduce this ?
>>>>>
>>>>> This bug is not found by the real execution.
>>>>> It is found by a static tool written by myself, and then I check
>>>>> it by manual code review.
>>>>
>>>> Before "line 2004: dev_kfree_skb_any(tx_skb->skb); ",
>>>>
>>>> "
>>>>
>>>> nv_disable_irq(dev);
>>>> nv_napi_disable(dev);
>>>> netif_tx_lock_bh(dev);
>>>> netif_addr_lock(dev);
>>>> spin_lock(&np->lock);
>>>> /* stop engines */
>>>> nv_stop_rxtx(dev); <---this stop rxtx
>>>> nv_txrx_reset(dev);
>>>> "
>>>>
>>>> In this case, does nv_start_xmit or nv_start_xmit_optimized still
>>>> work well?
>>>>
>>>
>>> nv_stop_rxtx() calls nv_stop_tx(dev).
>>>
>>> static void nv_stop_tx(struct net_device *dev)
>>> {
>>> struct fe_priv *np = netdev_priv(dev);
>>> u8 __iomem *base = get_hwbase(dev);
>>> u32 tx_ctrl = readl(base + NvRegTransmitterControl);
>>>
>>> if (!np->mac_in_use)
>>> tx_ctrl &= ~NVREG_XMITCTL_START;
>>> else
>>> tx_ctrl |= NVREG_XMITCTL_TX_PATH_EN;
>>> writel(tx_ctrl, base + NvRegTransmitterControl);
>>> if (reg_delay(dev, NvRegTransmitterStatus, NVREG_XMITSTAT_BUSY, 0,
>>> NV_TXSTOP_DELAY1, NV_TXSTOP_DELAY1MAX))
>>> netdev_info(dev, "%s: TransmitterStatus remained busy\n",
>>> __func__);
>>>
>>> udelay(NV_TXSTOP_DELAY2);
>>> if (!np->mac_in_use)
>>> writel(readl(base + NvRegTransmitPoll) &
>>> NVREG_TRANSMITPOLL_MAC_ADDR_REV,
>>> base + NvRegTransmitPoll);
>>> }
>>>
>>> nv_stop_tx() seems to only write registers to stop transmitting for
>>> hardware.
>>> But it does not wait until nv_start_xmit() and
>>> nv_start_xmit_optimized() finish execution.
>> There are 3 modes in forcedeth NIC.
>> In throughput mode (0), every tx & rx packet will generate an interrupt.
>> In CPU mode (1), interrupts are controlled by a timer.
>> In dynamic mode (2), the mode toggles between throughput and CPU mode
>> based on network load.
>>
>> From the source code,
>>
>> "np->recover_error = 1;" is related with CPU mode.
>>
>> nv_start_xmit or nv_start_xmit_optimized seems related with
>> ghroughput mode.
>>
>> In static void nv_do_nic_poll(struct timer_list *t),
>> when if (np->recover_error), line 2004:
>> dev_kfree_skb_any(tx_skb->skb); will run.
>>
>> When "np->recover_error=1", do you think nv_start_xmit or
>> nv_start_xmit_optimized will be called?
>
> Sorry, I do not know about these modes...
> But I still think nv_start_xmit() or nv_start_xmit_optimized() can be
> called here, in no matter which mode :)
:-P
If you have forcedeth NIC, you can make tests with it.:-)
>
>
> Best wishes,
> Jia-Ju Bai
>
On 2019/1/9 11:24, Yanjun Zhu wrote:
>
> If you have forcedeth NIC, you can make tests with it.:-)
Ah, I would like to, but I do not have the hardware...
Best wishes,
Jia-Ju Bai
From: Jia-Ju Bai <[email protected]>
Date: Tue, 8 Jan 2019 20:45:18 +0800
> In drivers/net/ethernet/nvidia/forcedeth.c, the functions
> nv_start_xmit() and nv_start_xmit_optimized() can be concurrently
> executed with nv_poll_controller().
>
> nv_start_xmit
> line 2321: prev_tx_ctx->skb = skb;
>
> nv_start_xmit_optimized
> line 2479: prev_tx_ctx->skb = skb;
>
> nv_poll_controller
> nv_do_nic_poll
> line 4134: spin_lock(&np->lock);
> nv_drain_rxtx
> nv_drain_tx
> nv_release_txskb
> line 2004: dev_kfree_skb_any(tx_skb->skb);
>
> Thus, two possible concurrency use-after-free bugs may occur.
I do not think so, the netif_tx_lock_bh() done will prevent the parallel
execution.