8139too: fix improper usage of workqueue
Subsystems (2)
PATCH 1/1 8139too: fix improper usage of workqueue
This driver was confused about how workqueue's work. It is save to schedule multiple times, and the flag (bitfield) was unnecessary and racy. The workqueue was never canceled either, so it is possible that the thread could run after driver was removes [OOPS]. As a bonus, get rid the other wasteful usage of bitfields.
Compile tested only, but based on some internal work done to fix reported oops.
drivers/net/8139too.c +7 -10
@@ -592,9 +592,8 @@ struct rtl8139_private {/* Twister tune state. */ char twistie, twist_row, twist_col; - unsigned int watchdog_fired : 1; - unsigned int default_port : 4; /* Last dev->if_port value. */ - unsigned int have_thread : 1; +unsigned char watchdog_fired; +unsigned char default_port;/* Last dev->if_port value. */ spinlock_t lock;spinlock_t rx_lock;
@@ -1614,8 +1613,7 @@ static void rtl8139_thread (struct work_ }else rtl8139_thread_iter (dev, tp, tp->mmio_addr); -if (tp->have_thread) -schedule_delayed_work (&tp->thread, thr_delay); +schedule_delayed_work (&tp->thread, thr_delay); out_unlock:rtnl_unlock (); }
@@ -1628,7 +1626,6 @@ static void rtl8139_start_thread(structelse if (tp->drv_flags &HAS_LNK_CHNG )return ; - tp->have_thread =1 ; tp->watchdog_fired =0 ;schedule_delayed_work (&tp->thread, next_tick);
@@ -1692,10 +1689,8 @@ static void rtl8139_tx_timeout (struct nstruct rtl8139_private *tp =netdev_priv (dev); tp->watchdog_fired =1 ; -if (!tp->have_thread) { -INIT_DELAYED_WORK (&tp->thread, rtl8139_thread); -schedule_delayed_work (&tp->thread, next_tick); - } + +schedule_delayed_work (&tp->thread, next_tick); }static int rtl8139_start_xmit (struct sk_buff *skb,struct net_device *dev)
@@ -2217,6 +2212,8 @@ static int rtl8139_close (struct net_devvoid __iomem *ioaddr = tp->mmio_addr;unsigned long flags; +cancel_delayed_work (&tp->thread); +netif_stop_queue (dev);napi_disable (&tp->napi);
Stephen Hemminger <[email protected]> :
Quoted 2 lines
This driver was confused about how workqueue's work. It is save to schedule multiple times, and the flag (bitfield) was unnecessary and racy.
static void rtl8139_start_thread(struct rtl8139_private *tp)
{
tp->twistie = 0;
if (tp->chipset == CH_8139_K)
tp->twistie = 1;
else if (tp->drv_flags & HAS_LNK_CHNG)
return;
^^^^^^^
tp->have_thread = 1;The use of the thread was dependent on RTL8129_CAPS. Now it will be scheduled in the tx_timeout handler for RTL8139_CAPS too.
Speaking of race, the bitfield ('have_thread') is initialized through rtl8139_open
-> rtl8139_start_thread
Do you mean that the tx_timeout handler can kick before the bitfield is set ?
Quoted 2 lines
The workqueue was never canceled either, so it is possible that the thread could run after driver was removes [OOPS].
The driver uses flush_scheduled_work() when the pci device is removed.
Where would the workqueue be scheduled after / during the removal ?
Signature · 1 line
Ueimor
Quoted 4 lines
The driver uses
flush_scheduled_work()when the pci device is removed.Where would the workqueue be scheduled after / during the removal ?
The timer can fire after removal causing workqueue to be scheduled. flush_scheduled_work only impacts work already scheduled not delayed.
How about this (simpler) version.
drivers/net/8139too.c +7 -6
@@ -592,9 +592,9 @@ struct rtl8139_private {/* Twister tune state. */ char twistie, twist_row, twist_col; - unsigned int watchdog_fired : 1; - unsigned int default_port : 4; /* Last dev->if_port value. */ - unsigned int have_thread : 1; +bool watchdog_fired; +bool have_thread; +u8 default_port;/* Last dev->if_port value. */ spinlock_t lock;spinlock_t rx_lock;
@@ -1692,10 +1692,9 @@ static void rtl8139_tx_timeout (struct nstruct rtl8139_private *tp =netdev_priv (dev); tp->watchdog_fired =1 ; -if (!tp->have_thread) { -INIT_DELAYED_WORK (&tp->thread, rtl8139_thread); + +if (!tp->have_thread)schedule_delayed_work (&tp->thread, next_tick); - } }static int rtl8139_start_xmit (struct sk_buff *skb,struct net_device *dev)
@@ -2217,6 +2216,8 @@ static int rtl8139_close (struct net_devvoid __iomem *ioaddr = tp->mmio_addr;unsigned long flags; +cancel_delayed_work (&tp->thread); +netif_stop_queue (dev);napi_disable (&tp->napi);
Stephen Hemminger <[email protected]> :
[...]
Quoted 1 line
flush_scheduled_workonly impacts work already scheduled not delayed.
Ok. I was confused.
Quoted 1 line
How about this (simpler) version.
Signature · 1 line
Ueimor