[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <7af42c1e-8a18-407a-8abb-a43a1362765f@intel.com>
Date: Tue, 7 Jan 2025 09:55:13 +0200
From: "Lifshits, Vitaly" <vitaly.lifshits@...el.com>
To: Dmitrii Ermakov <demonihin@...il.com>, <davem@...emloft.net>
CC: <edumazet@...gle.com>, <kuba@...nel.org>, <pabeni@...hat.com>,
<przemyslaw.kitszel@...el.com>, <anthony.l.nguyen@...el.com>,
<netdev@...r.kernel.org>, <intel-wired-lan@...ts.osuosl.org>,
<linux-kernel@...r.kernel.org>
Subject: Re: [Intel-wired-lan] [PATCH v2 RESEND net-next] e1000e: makes
e1000_watchdog_task use queue_delayed_work
On 1/5/2025 1:38 PM, Dmitrii Ermakov wrote:
> Replaces watchdog timer with delayed_work as advised
> in the driver's TODO comment.
>
> Signed-off-by: Dmitrii Ermakov <demonihin@...il.com>
> ---
> V1 -> V2: Removed redundant line wraps, renamed e1000_watchdog to e1000_watchdog_work
>
> drivers/net/ethernet/intel/e1000e/e1000.h | 4 +--
> drivers/net/ethernet/intel/e1000e/netdev.c | 42 ++++++++--------------
> 2 files changed, 16 insertions(+), 30 deletions(-)
>
> diff --git a/drivers/net/ethernet/intel/e1000e/e1000.h b/drivers/net/ethernet/intel/e1000e/e1000.h
> index ba9c19e6994c..5a60372d2158 100644
> --- a/drivers/net/ethernet/intel/e1000e/e1000.h
> +++ b/drivers/net/ethernet/intel/e1000e/e1000.h
> @@ -189,12 +189,12 @@ struct e1000_phy_regs {
>
> /* board specific private data structure */
> struct e1000_adapter {
> - struct timer_list watchdog_timer;
> struct timer_list phy_info_timer;
> struct timer_list blink_timer;
>
> + struct delayed_work watchdog_work;
> +
> struct work_struct reset_task;
> - struct work_struct watchdog_task;
>
> const struct e1000_info *ei;
>
> diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c b/drivers/net/ethernet/intel/e1000e/netdev.c
> index 286155efcedf..cb68662cdc3a 100644
> --- a/drivers/net/ethernet/intel/e1000e/netdev.c
> +++ b/drivers/net/ethernet/intel/e1000e/netdev.c
> @@ -1778,7 +1778,7 @@ static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
> }
> /* guard against interrupt when we're going down */
> if (!test_bit(__E1000_DOWN, &adapter->state))
> - mod_timer(&adapter->watchdog_timer, jiffies + 1);
> + queue_delayed_work(system_wq, &adapter->watchdog_work, 1);
> }
>
> /* Reset on uncorrectable ECC error */
> @@ -1857,7 +1857,7 @@ static irqreturn_t e1000_intr(int __always_unused irq, void *data)
> }
> /* guard against interrupt when we're going down */
> if (!test_bit(__E1000_DOWN, &adapter->state))
> - mod_timer(&adapter->watchdog_timer, jiffies + 1);
> + queue_delayed_work(system_wq, &adapter->watchdog_work, 1);
> }
>
> /* Reset on uncorrectable ECC error */
> @@ -1901,7 +1901,7 @@ static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
> hw->mac.get_link_status = true;
> /* guard against interrupt when we're going down */
> if (!test_bit(__E1000_DOWN, &adapter->state))
> - mod_timer(&adapter->watchdog_timer, jiffies + 1);
> + queue_delayed_work(system_wq, &adapter->watchdog_work, 1);
> }
>
> if (!test_bit(__E1000_DOWN, &adapter->state))
> @@ -4287,7 +4287,8 @@ void e1000e_down(struct e1000_adapter *adapter, bool reset)
>
> napi_synchronize(&adapter->napi);
>
> - del_timer_sync(&adapter->watchdog_timer);
> + cancel_delayed_work_sync(&adapter->watchdog_work);
> +
> del_timer_sync(&adapter->phy_info_timer);
>
> spin_lock(&adapter->stats64_lock);
> @@ -5169,25 +5170,12 @@ static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
> }
> }
>
> -/**
> - * e1000_watchdog - Timer Call-back
> - * @t: pointer to timer_list containing private info adapter
> - **/
> -static void e1000_watchdog(struct timer_list *t)
> +static void e1000_watchdog_work(struct work_struct *work)
> {
> - struct e1000_adapter *adapter = from_timer(adapter, t, watchdog_timer);
> -
> - /* Do the rest outside of interrupt context */
> - schedule_work(&adapter->watchdog_task);
> -
> - /* TODO: make this use queue_delayed_work() */
> -}
> -
> -static void e1000_watchdog_task(struct work_struct *work)
> -{
> - struct e1000_adapter *adapter = container_of(work,
> - struct e1000_adapter,
> - watchdog_task);
> + struct delayed_work *dwork =
> + container_of(work, struct delayed_work, work);
> + struct e1000_adapter *adapter =
> + container_of(dwork, struct e1000_adapter, watchdog_work);
> struct net_device *netdev = adapter->netdev;
> struct e1000_mac_info *mac = &adapter->hw.mac;
> struct e1000_phy_info *phy = &adapter->hw.phy;
> @@ -5416,8 +5404,8 @@ static void e1000_watchdog_task(struct work_struct *work)
>
> /* Reset the timer */
> if (!test_bit(__E1000_DOWN, &adapter->state))
> - mod_timer(&adapter->watchdog_timer,
> - round_jiffies(jiffies + 2 * HZ));
> + queue_delayed_work(system_wq, &adapter->watchdog_work,
> + round_jiffies(2 * HZ));
> }
>
> #define E1000_TX_FLAGS_CSUM 0x00000001
> @@ -7596,11 +7584,10 @@ static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
> goto err_eeprom;
> }
>
> - timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
> timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
> + INIT_DELAYED_WORK(&adapter->watchdog_work, e1000_watchdog_work);
>
> INIT_WORK(&adapter->reset_task, e1000_reset_task);
> - INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
> INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
> INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
> INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
> @@ -7741,11 +7728,10 @@ static void e1000_remove(struct pci_dev *pdev)
> * from being rescheduled.
> */
> set_bit(__E1000_DOWN, &adapter->state);
> - del_timer_sync(&adapter->watchdog_timer);
> + cancel_delayed_work_sync(&adapter->watchdog_work);
> del_timer_sync(&adapter->phy_info_timer);
>
> cancel_work_sync(&adapter->reset_task);
> - cancel_work_sync(&adapter->watchdog_task);
> cancel_work_sync(&adapter->downshift_task);
> cancel_work_sync(&adapter->update_phy_task);
> cancel_work_sync(&adapter->print_hang_task);
Hi Dmitrii,
I have found that in the past someone has already tried to change
delayed work instead of watchdog task (59653e6497d1: e1000e: Make
watchdog use delayed work). This resulted in driver crashes and
connections to be reset unexpectedly (d5ad7a6a7f3c8: e1000e: Revert
"e1000e: Make watchdog use delayed work").
Because of that, and unless there is a clear benefit to using
delayed_work, I recommend to reject this patch, as the risk of
regression is high.
Powered by blists - more mailing lists