Created attachment 17534 [details]
dmesg log

Reply-To: akpm@linux-foundation.org

On Fri, 29 Aug 2008 14:08:09 -0700 (PDT)
bugme-daemon@bugzilla.kernel.org wrote:

> http://bugzilla.kernel.org/show_bug.cgi?id=11454
> 
>            Summary: atl1e - BUG: scheduling while atomic:
>                     modprobe/678/0x00000002
>            Product: Drivers
>            Version: 2.5
>      KernelVersion: 2.6.27-rc5
>           Platform: All
>         OS/Version: Linux
>               Tree: Mainline
>             Status: NEW
>           Severity: high
>           Priority: P1
>          Component: Network
>         AssignedTo: jgarzik@pobox.com
>         ReportedBy: harn-solo@gmx.de
> 
> 
> Latest working kernel version: - 
> Earliest failing kernel version: 2.6.27-rc1
> Software: x86_64
> 
> ...
> 
> Problem Description: Loading the module atl1e results in the following
> BUG-trace:
> 
> [    7.657410] ATL1E 0000:02:00.0: PCI INT A -> GSI 17 (level, low) -> IRQ 17
> [    7.657421] ATL1E 0000:02:00.0: setting latency timer to 64
> [    7.660012] BUG: scheduling while atomic: modprobe/678/0x00000002
> [    7.660014] Modules linked in: atl1e(+) sg thermal i2c_core evdev rtc
> processor button
> [    7.660021] Pid: 678, comm: modprobe Not tainted 2.6.27-rc5 #1
> [    7.660023] 
> [    7.660024] Call Trace:
> [    7.660030]  [<ffffffff8022f7aa>] __schedule_bug+0x62/0x67
> [    7.660034]  [<ffffffff80402606>] schedule+0xdb/0x84c
> [    7.660037]  [<ffffffff80402db4>] ? thread_return+0x3d/0xcb
> [    7.660041]  [<ffffffff8023bfb8>] ? __mod_timer+0xbe/0xd0
> [    7.660044]  [<ffffffff804030df>] schedule_timeout+0x8d/0xb4
> [    7.660047]  [<ffffffff8023b94c>] ? process_timeout+0x0/0xb
> [    7.660049]  [<ffffffff804030da>] ? schedule_timeout+0x88/0xb4
> [    7.660052]  [<ffffffff8040311f>]
> schedule_timeout_uninterruptible+0x19/0x1b
> [    7.660055]  [<ffffffff8023bfde>] msleep+0x14/0x1e
> [    7.660061]  [<ffffffffa002df2a>] atl1e_phy_init+0x53/0x30f [atl1e]
> [    7.660066]  [<ffffffffa002f26c>] atl1e_probe+0x534/0x690 [atl1e]
> [    7.660070]  [<ffffffff80302529>] ? kobject_get+0x1a/0x22
> [    7.660073]  [<ffffffff803110f2>] pci_device_probe+0x4f/0x75
> [    7.660077]  [<ffffffff8036973a>] driver_probe_device+0xc0/0x16e
> [    7.660080]  [<ffffffff80369837>] __driver_attach+0x4f/0x79
> [    7.660082]  [<ffffffff803697e8>] ? __driver_attach+0x0/0x79
> [    7.660085]  [<ffffffff80368fcf>] bus_for_each_dev+0x4f/0x89
> [    7.660087]  [<ffffffff80302529>] ? kobject_get+0x1a/0x22
> [    7.660090]  [<ffffffff80369585>] driver_attach+0x1c/0x1e
> [    7.660092]  [<ffffffff803688ef>] bus_add_driver+0xb7/0x201
> [    7.660095]  [<ffffffff80369a28>] driver_register+0xa8/0x128
> [    7.660097]  [<ffffffff8031136f>] __pci_register_driver+0x66/0x9f
> [    7.660102]  [<ffffffffa0015000>] ? atl1e_init_module+0x0/0x20 [atl1e]
> [    7.660107]  [<ffffffffa001501e>] atl1e_init_module+0x1e/0x20 [atl1e]
> [    7.660110]  [<ffffffff80209047>] _stext+0x47/0x13f
> [    7.660113]  [<ffffffff802562ca>] sys_init_module+0xa9/0x1b6
> [    7.660116]  [<ffffffff8020b5bb>] system_call_fastpath+0x16/0x1b
> [    7.660118] 
> [    7.663008] BUG: scheduling while atomic: modprobe/678/0x00000002

Guys, this is fairly pathetic.

	spin_lock(&adapter->mdio_lock);
	atl1e_phy_init(&adapter->hw);
	spin_unlock(&adapter->mdio_lock);

int atl1e_phy_init(struct atl1e_hw *hw)
{
	...
	msleep(...);
	...
	msleep(...);
	...
	msleep(...);
	...
	msleep(...);
}

we spent all that time developing all that runtime-debugging
infrastructure and many people are just ignoring it.

Please, read Documentation/SubmitChecklist very carefully and *do what
it says*.

Please also fix the above bug.


btw, these casts:

drivers/net/atl1e/atl1e_hw.c:   struct atl1e_adapter *adapter = (struct atl1e_adapter *)hw->adapter;
drivers/net/atl1e/atl1e_hw.c:   struct atl1e_adapter *adapter = (struct atl1e_adapter *)hw->adapter;
drivers/net/atl1e/atl1e_hw.c:   struct atl1e_adapter *adapter = (struct atl1e_adapter *)hw->adapter;

are unneeded and undesirable.  hw->adapter already has type atl1e_adapter*.

Reply-To: jie.yang@Atheros.com

from Jie Yang <jie.yang@atheros.com>

 On Saturday, August 30, 2008 5:27 AM
 Andrew Morton <akpm@linux-foundation.org]>
> On Fri, 29 Aug 2008 14:08:09 -0700 (PDT) 
> bugme-daemon@bugzilla.kernel.org wrote:
> 
> > http://bugzilla.kernel.org/show_bug.cgi?id=11454
> >
> >            Summary: atl1e - BUG: scheduling while atomic:
> >                     modprobe/678/0x00000002
> >            Product: Drivers
> >            Version: 2.5
> >      KernelVersion: 2.6.27-rc5
> >           Platform: All
> >         OS/Version: Linux
> >               Tree: Mainline
> >             Status: NEW
> >           Severity: high
> >           Priority: P1
> >          Component: Network
> >         AssignedTo: jgarzik@pobox.com
> >         ReportedBy: harn-solo@gmx.de
> >
> >
> > Latest working kernel version: -
> > Earliest failing kernel version: 2.6.27-rc1
> > Software: x86_64
> >
> 
> drivers/net/atl1e/atl1e_hw.c:   struct atl1e_adapter *adapter 
> = (struct atl1e_adapter *)hw->adapter;
> drivers/net/atl1e/atl1e_hw.c:   struct atl1e_adapter *adapter 
> = (struct atl1e_adapter *)hw->adapter;
> drivers/net/atl1e/atl1e_hw.c:   struct atl1e_adapter *adapter 
> = (struct atl1e_adapter *)hw->adapter;
> 
> are unneeded and undesirable.  hw->adapter already has type 
> atl1e_adapter*.
> 

just as Matthew Wilcox <matthew@wil.cx> mentioned:
> Lockdep warns about the mdio_lock taken with interrupts enabled then 
> later taken from interrupt context.  Initially, I considered changing 
> these to spin_lock_irq/spin_unlock_irq, but then I looked at 
> atl1e_phy_init() and saw that it calls msleep().  Sleeping while 
> holding a spinlock is not allowed either.
> 
> In the probe path, we haven't registered the interrupt handler, so it 
> can't poke at this card yet.  It's before we call register_netdev(), 
> so I don't think any other threads can reach this card either.  If I'm 
> right, we don't need a spinlock at all.

So, just do not take mdio_lock lock in atl1e_probe, and remove the
unneeded (struct atl1e_adapter *)

Signed-off-by: Jie Yang <jie.yang@atheros.com>
---

BTW: I do not know if this format is suitable for repling [Bugme-new],
if it is not suitable, just let me know.

diff --git a/drivers/net/atl1e/atl1e_hw.c b/drivers/net/atl1e/atl1e_hw.c
index 949e753..8cbc1b5 100644
--- a/drivers/net/atl1e/atl1e_hw.c
+++ b/drivers/net/atl1e/atl1e_hw.c
@@ -397,7 +397,7 @@ static int atl1e_phy_setup_autoneg_adv(struct atl1e_hw *hw)
  */
 int atl1e_phy_commit(struct atl1e_hw *hw)
 {
-       struct atl1e_adapter *adapter = (struct atl1e_adapter *)hw->adapter;
+       struct atl1e_adapter *adapter = hw->adapter;
        struct pci_dev *pdev = adapter->pdev;
        int ret_val;
        u16 phy_data;
@@ -431,7 +431,7 @@ int atl1e_phy_commit(struct atl1e_hw *hw)

 int atl1e_phy_init(struct atl1e_hw *hw)
 {
-       struct atl1e_adapter *adapter = (struct atl1e_adapter *)hw->adapter;
+       struct atl1e_adapter *adapter = hw->adapter;
        struct pci_dev *pdev = adapter->pdev;
        s32 ret_val;
        u16 phy_val;
@@ -525,7 +525,7 @@ int atl1e_phy_init(struct atl1e_hw *hw)
  */
 int atl1e_reset_hw(struct atl1e_hw *hw)
 {
-       struct atl1e_adapter *adapter = (struct atl1e_adapter *)hw->adapter;
+       struct atl1e_adapter *adapter = hw->adapter;
        struct pci_dev *pdev = adapter->pdev;

        u32 idle_status_data = 0;
diff --git a/drivers/net/atl1e/atl1e_main.c b/drivers/net/atl1e/atl1e_main.c
index 7685b99..9b60352 100644
--- a/drivers/net/atl1e/atl1e_main.c
+++ b/drivers/net/atl1e/atl1e_main.c
@@ -2390,9 +2390,7 @@ static int __devinit atl1e_probe(struct pci_dev *pdev,
        }

        /* Init GPHY as early as possible due to power saving issue  */
-       spin_lock(&adapter->mdio_lock);
        atl1e_phy_init(&adapter->hw);
-       spin_unlock(&adapter->mdio_lock);
        /* reset the controller to
         * put the device in a known good starting state */
        err = atl1e_reset_hw(&adapter->hw);

On Sun, Aug 31, 2008 at 12:51:58PM +0800, jie.yang@atheros.com wrote:
> from Jie Yang <jie.yang@atheros.com>
> 
>  On Saturday, August 30, 2008 5:27 AM
>  Andrew Morton <akpm@linux-foundation.org]>
> > On Fri, 29 Aug 2008 14:08:09 -0700 (PDT) 
> > bugme-daemon@bugzilla.kernel.org wrote:
> > 
> > > http://bugzilla.kernel.org/show_bug.cgi?id=11454
> > >
> > >            Summary: atl1e - BUG: scheduling while atomic:
> > >                     modprobe/678/0x00000002
> > >            Product: Drivers
> > >            Version: 2.5
> > >      KernelVersion: 2.6.27-rc5
> > >           Platform: All
> > >         OS/Version: Linux
> > >               Tree: Mainline
> > >             Status: NEW
> > >           Severity: high
> > >           Priority: P1
> > >          Component: Network
> > >         AssignedTo: jgarzik@pobox.com

How about taking my original patch, sent August 12th instead?  That has
a good changelog and proper attribution.  The removal of the
unnecessary casts can be a separate message.

If you weren't cc'd on the patch (Jeff was), you can pick it up from
netdev.

Reply-To: Jie.Yang@Atheros.com

On Sunday, August 31, 2008 1:09 PM
Matthew Wilcox <matthew@wil.cx> wrote:
> How about taking my original patch, sent August 12th instead?
>  That has a good changelog and proper attribution.  The
> removal of the unnecessary casts can be a separate message.
>
> If you weren't cc'd on the patch (Jeff was), you can pick it
> up from netdev.
>
Yes, I do have the patch, should I resend with Signed-off-by: Matthew Wilcox <willy@linux.intel.com>.
for that patch may have conflicts now.

the origin patch:
diff --git a/drivers/net/atl1e/atl1e_main.c b/drivers/net/atl1e/atl1e_main.c index 82d7be1..ba22a51 100644
--- a/drivers/net/atl1e/atl1e_main.c
+++ b/drivers/net/atl1e/atl1e_main.c
@@ -2389,9 +2389,7 @@ static int __devinit atl1e_probe(struct pci_dev *pdev,
        }

        /* Init GPHY as early as possible due to power saving issue  */
-       spin_lock(&adapter->mdio_lock);
        atl1e_phy_init(&adapter->hw);
-       spin_unlock(&adapter->mdio_lock);
        /* reset the controller to
         * put the device in a known good starting state */
        err = atl1e_reset_hw(&adapter->hw);

the new one:
diff --git a/drivers/net/atl1e/atl1e_main.c b/drivers/net/atl1e/atl1e_main.c
index 7685b99..9b60352 100644
--- a/drivers/net/atl1e/atl1e_main.c
+++ b/drivers/net/atl1e/atl1e_main.c
@@ -2390,9 +2390,7 @@ static int __devinit atl1e_probe(struct pci_dev *pdev,
        }

        /* Init GPHY as early as possible due to power saving issue  */
-       spin_lock(&adapter->mdio_lock);
        atl1e_phy_init(&adapter->hw);
-       spin_unlock(&adapter->mdio_lock);
        /* reset the controller to
         * put the device in a known good starting state */
        err = atl1e_reset_hw(&adapter->hw);

the line num changed from 2389 to 2390.  :(

Best wishes
jie

*** Bug 11510 has been marked as a duplicate of this bug. ***

Since 2.6.27-rc6 (rc8 also tested), the LAN-adapter works great for me. Although I'm still getting two of those "BUG: scheduling while atomic" traces in dmesg.

Thanks.

As for 2.6.27.2 it is fixed now like described in the patch changelog.

Fixed for me, thanks