Re: BUG in latest GIT 2.6.28-0658-g637b180

[Bartlomiej Zolnierkiewicz]

On Wednesday 07 January 2009, Larry Finger wrote:
> The latest GIT generates the following BUG.
> 
> 
> BUG: sleeping function called from invalid context at mm/slab.c:3046
> in_atomic(): 0, irqs_disabled(): 1, pid: 523, name: modprobe
> no locks held by modprobe/523.
> irq event stamp: 17802
> hardirqs last  enabled at (17801): [<ffffffff8045dd99>]
> _spin_unlock_irqrestore+0x44/0x4c
> hardirqs last disabled at (17802): [<ffffffffa00a4792>]
> ide_probe_port+0x6c/0x469 [ide_core]
> softirqs last  enabled at (17796): [<ffffffff802431f3>]
> __do_softirq+0x159/0x166
> softirqs last disabled at (17789): [<ffffffff8020d4bc>]
> call_softirq+0x1c/0x28
> Pid: 523, comm: modprobe Not tainted 2.6.28-Linus-06858-g637b180 #31
> Call Trace:
>  [<ffffffff8025dbdd>] ? print_irqtrace_events+0xd0/0xd4
>  [<ffffffff80236bcd>] __might_sleep+0xf9/0xfb
>  [<ffffffff802b5469>] kmem_cache_alloc+0x2d/0x142
>  [<ffffffffa00a489b>] ide_probe_port+0x175/0x469 [ide_core]
>  [<ffffffffa00a5750>] ide_host_register+0x2b8/0x611 [ide_core]
>  [<ffffffffa00a8526>] ? do_ide_setup_pci_device+0x70/0xd3 [ide_core]
>  [<ffffffffa00a8f1d>] ide_pci_init_one+0xcc/0xee [ide_core]
>  [<ffffffff8025e852>] ? trace_hardirqs_off+0xd/0xf
>  [<ffffffff8045dd92>] ? _spin_unlock_irqrestore+0x3d/0x4c
>  [<ffffffff8023e3ec>] ? release_console_sem+0x1b0/0x1e5
>  [<ffffffff8045aac8>] ? printk+0x41/0x49
>  [<ffffffffa00bf604>] amd74xx_probe+0x13e/0x145 [amd74xx]
>  [<ffffffffa00bf340>] ? init_chipset_amd74xx+0x0/0x160 [amd74xx]
>  [<ffffffffa00bf4a0>] ? init_hwif_amd74xx+0x0/0x26 [amd74xx]
>  [<ffffffff80345fb6>] pci_device_probe+0xc6/0x11d
>  [<ffffffff803b7f1d>] driver_probe_device+0xc3/0x168
>  [<ffffffff803b801f>] __driver_attach+0x5d/0x80
>  [<ffffffff803b7fc2>] ? __driver_attach+0x0/0x80
>  [<ffffffff803b7794>] bus_for_each_dev+0x59/0x90
>  [<ffffffff803b7d60>] driver_attach+0x21/0x23
>  [<ffffffff803b703b>] bus_add_driver+0xbf/0x20c
>  [<ffffffff803b822e>] driver_register+0xb0/0x130
>  [<ffffffff8034628d>] __pci_register_driver+0x68/0xa3
>  [<ffffffffa0023000>] ? amd74xx_ide_init+0x0/0x20 [amd74xx]
>  [<ffffffffa002301e>] amd74xx_ide_init+0x1e/0x20 [amd74xx]
>  [<ffffffff8020905b>] _stext+0x5b/0x135
>  [<ffffffff802561b0>] ? __blocking_notifier_call_chain+0x5d/0x6f
>  [<ffffffff8026af0c>] sys_init_module+0xb0/0x1d3
>  [<ffffffff8020c2cb>] system_call_fastpath+0x16/0x1b

Fix below.  Thanks again Larry!

[ I now always run with LOCKDEP=y but the issue was probably hidden by the
  ide_probe_port() IRQ masking cleanup (http://lkml.org/lkml/2009/1/6/267)
  so I didn't catch it before push to Linus... ]

From: Bartlomiej Zolnierkiewicz <bzolnier@xxxxxxxxx>
Subject: [PATCH] ide: fix accidental LOCKDEP breakage caused by local_irq_set() removal

commit 54cc1428cfa619e16d75baae8cb041a2eff015f0 ("ide: remove
local_irq_set() macro") accidentally replaced local_save_flags()
by local_irq_set() in ide_probe_port() and __ide_wait_stat()
which resulted in LOCKDEP breakage.

Reported-by: Larry Finger <Larry.Finger@xxxxxxxxxxxx>
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@xxxxxxxxx>
---
 drivers/ide/ide-iops.c  |    2 +-
 drivers/ide/ide-probe.c |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

Index: b/drivers/ide/ide-iops.c
===================================================================
--- a/drivers/ide/ide-iops.c
+++ b/drivers/ide/ide-iops.c
@@ -493,7 +493,7 @@ static int __ide_wait_stat(ide_drive_t *
 	stat = tp_ops->read_status(hwif);
 
 	if (stat & ATA_BUSY) {
-		local_irq_save(flags);
+		local_save_flags(flags);
 		local_irq_enable_in_hardirq();
 		timeout += jiffies;
 		while ((stat = tp_ops->read_status(hwif)) & ATA_BUSY) {
Index: b/drivers/ide/ide-probe.c
===================================================================
--- a/drivers/ide/ide-probe.c
+++ b/drivers/ide/ide-probe.c
@@ -796,7 +796,7 @@ static int ide_probe_port(ide_hwif_t *hw
 	if (irqd)
 		disable_irq(hwif->irq);
 
-	local_irq_save(flags);
+	local_save_flags(flags);
 	local_irq_enable_in_hardirq();
 
 	if (ide_port_wait_ready(hwif) == -EBUSY)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/