Change since v1:
- restored __exit for module_exit function
---8<---
>From c04a24acebee4b4432afa7a2a7d4cfe114cc9a40 Mon Sep 17 00:00:00 2001
From: Corentin Labbe <[email protected]>
Date: Mon, 17 Oct 2016 14:17:49 +0200
Subject: [PATCH v2] rtc: cmos: remove all __exit_p annotations
I got the following stack trace under qemu:
[ 7.575243] BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
[ 7.596098] IP: [<ffffffff814f5b08>] cmos_set_alarm+0x38/0x280
[ 7.615699] PGD 3ccbe067
[ 7.615923] PUD 3daf2067
[ 7.635156] PMD 0
[ 7.654358] Oops: 0000 [#1] SMP
[ 7.673869] Modules linked in:
[ 7.693235] CPU: 0 PID: 1701 Comm: hwclock Tainted: G W 4.9.0-rc1+ #24
[ 7.712455] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.9.3-0-ge2fc41e-prebuilt.qemu-project.org 04/01/2014
[ 7.753569] task: ffff88003d88dc40 task.stack: ffffc90000224000
[ 7.773743] RIP: 0010:[<ffffffff814f5b08>] [<ffffffff814f5b08>] cmos_set_alarm+0x38/0x280
[ 7.794893] RSP: 0018:ffffc90000227c10 EFLAGS: 00010296
[ 7.815890] RAX: 000000000000001d RBX: ffffc90000227d28 RCX: ffffffff8182be78
[ 7.836057] RDX: 0000000000000001 RSI: 0000000000000202 RDI: 0000000000000202
[ 7.856612] RBP: ffffc90000227c48 R08: 0000000000000000 R09: 0000000000000001
[ 7.877561] R10: 00000000000001c0 R11: 00000000000001c0 R12: 0000000000000000
[ 7.897072] R13: ffff88003d96f400 R14: ffff88003dac6410 R15: ffff88003dac6420
[ 7.917403] FS: 00007f77f42d9700(0000) GS:ffff88003fc00000(0000) knlGS:0000000000000000
[ 7.938293] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 7.958364] CR2: 0000000000000010 CR3: 000000003ccbb000 CR4: 00000000000006f0
[ 7.978028] Stack:
[ 7.997120] ffff88003dac6000 ffff88003dac6410 0000000058049d01 ffffc90000227d28
[ 8.016993] ffff88003dac6000 ffff88003dac6410 ffff88003dac6420 ffffc90000227c98
[ 8.039505] ffffffff814f225d 0000001800227c98 000000090000002a 0000000900000011
[ 8.059985] Call Trace:
[ 8.080110] [<ffffffff814f225d>] __rtc_set_alarm+0x8d/0xa0
[ 8.099421] [<ffffffff814f2389>] rtc_timer_enqueue+0x119/0x190
[ 8.119925] [<ffffffff814f2e6e>] rtc_update_irq_enable+0xbe/0x100
[ 8.140583] [<ffffffff814f3bb0>] rtc_dev_ioctl+0x3c0/0x480
[ 8.161162] [<ffffffff81146b6a>] ? user_path_at_empty+0x3a/0x50
[ 8.182717] [<ffffffff8114aa36>] do_vfs_ioctl+0x96/0x5c0
[ 8.204624] [<ffffffff8113e066>] ? vfs_stat+0x16/0x20
[ 8.225994] [<ffffffff8113e135>] ? SyS_newstat+0x15/0x30
[ 8.247043] [<ffffffff8114afa7>] SyS_ioctl+0x47/0x80
[ 8.267191] [<ffffffff815f5c77>] entry_SYSCALL_64_fastpath+0x1a/0xa9
[ 8.288719] Code: 6a 81 48 89 e5 41 57 41 56 41 55 49 89 fd 41 54 53 48 89 f3 48 c7 c6 20 c4 78 81 48 83 ec 10 e8 8f 00 ef ff 4d 8b a5 a0 00 00 00 <41> 8b 44 24 10 85 c0 0f 8e 2b 02 00 00 4c 89 ef 31 c0 b9 53 01
[ 8.335233] RIP [<ffffffff814f5b08>] cmos_set_alarm+0x38/0x280
[ 8.357096] RSP <ffffc90000227c10>
[ 8.379051] CR2: 0000000000000010
[ 8.401736] ---[ end trace 5cbcd83a1f225ed3 ]---
This occur only when CONFIG_DEBUG_TEST_DRIVER_REMOVE is enabled and
CONFIG_RTC_DRV_CMOS builtin.
When cmos_set_alarm() is called dev is NULL and so trigger the deref via
cmos->irq
The problem comes from that the device is removed but no remove function
are called due to _exit_p().
This patch remove all _exit_p() annotation.
Signed-off-by: Corentin Labbe <[email protected]>
---
drivers/rtc/rtc-cmos.c | 10 +++++-----
1 file changed, 5 insertions(+), 5 deletions(-)
diff --git a/drivers/rtc/rtc-cmos.c b/drivers/rtc/rtc-cmos.c
index dd3d598..6f0e12e 100644
--- a/drivers/rtc/rtc-cmos.c
+++ b/drivers/rtc/rtc-cmos.c
@@ -776,7 +776,7 @@ static void cmos_do_shutdown(int rtc_irq)
spin_unlock_irq(&rtc_lock);
}
-static void __exit cmos_do_remove(struct device *dev)
+static void cmos_do_remove(struct device *dev)
{
struct cmos_rtc *cmos = dev_get_drvdata(dev);
struct resource *ports;
@@ -1129,7 +1129,7 @@ static int cmos_pnp_probe(struct pnp_dev *pnp, const struct pnp_device_id *id)
pnp_irq(pnp, 0));
}
-static void __exit cmos_pnp_remove(struct pnp_dev *pnp)
+static void cmos_pnp_remove(struct pnp_dev *pnp)
{
cmos_do_remove(&pnp->dev);
}
@@ -1161,7 +1161,7 @@ static struct pnp_driver cmos_pnp_driver = {
.name = (char *) driver_name,
.id_table = rtc_ids,
.probe = cmos_pnp_probe,
- .remove = __exit_p(cmos_pnp_remove),
+ .remove = cmos_pnp_remove,
.shutdown = cmos_pnp_shutdown,
/* flag ensures resume() gets called, and stops syslog spam */
@@ -1238,7 +1238,7 @@ static int __init cmos_platform_probe(struct platform_device *pdev)
return cmos_do_probe(&pdev->dev, resource, irq);
}
-static int __exit cmos_platform_remove(struct platform_device *pdev)
+static int cmos_platform_remove(struct platform_device *pdev)
{
cmos_do_remove(&pdev->dev);
return 0;
@@ -1263,7 +1263,7 @@ static void cmos_platform_shutdown(struct platform_device *pdev)
MODULE_ALIAS("platform:rtc_cmos");
static struct platform_driver cmos_platform_driver = {
- .remove = __exit_p(cmos_platform_remove),
+ .remove = cmos_platform_remove,
.shutdown = cmos_platform_shutdown,
.driver = {
.name = driver_name,
--
2.7.3
Hi,
On 18/10/2016 at 13:55:53 +0200, Corentin Labbe wrote :
> Change since v1:
> - restored __exit for module_exit function
>
> ---8<---
I don't think this is the proper scissor symbol for git. At least, it is
no recognized by patchwork and messes your patch.
> From c04a24acebee4b4432afa7a2a7d4cfe114cc9a40 Mon Sep 17 00:00:00 2001
> From: Corentin Labbe <[email protected]>
> Date: Mon, 17 Oct 2016 14:17:49 +0200
> Subject: [PATCH v2] rtc: cmos: remove all __exit_p annotations
>
> I got the following stack trace under qemu:
[...]
> When cmos_set_alarm() is called dev is NULL and so trigger the deref via
> cmos->irq
>
> The problem comes from that the device is removed but no remove function
> are called due to _exit_p().
>
> This patch remove all _exit_p() annotation.
>
> Signed-off-by: Corentin Labbe <[email protected]>
> ---
The changelog should go here.
> drivers/rtc/rtc-cmos.c | 10 +++++-----
> 1 file changed, 5 insertions(+), 5 deletions(-)
>
--
Alexandre Belloni, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com
On Tue, Oct 18, 2016 at 04:16:39PM +0200, Alexandre Belloni wrote:
> Hi,
>
> On 18/10/2016 at 13:55:53 +0200, Corentin Labbe wrote :
> > Change since v1:
> > - restored __exit for module_exit function
> >
> > ---8<---
>
> I don't think this is the proper scissor symbol for git. At least, it is
> no recognized by patchwork and messes your patch.
>
> > From c04a24acebee4b4432afa7a2a7d4cfe114cc9a40 Mon Sep 17 00:00:00 2001
> > From: Corentin Labbe <[email protected]>
> > Date: Mon, 17 Oct 2016 14:17:49 +0200
> > Subject: [PATCH v2] rtc: cmos: remove all __exit_p annotations
> >
> > I got the following stack trace under qemu:
>
> [...]
> > When cmos_set_alarm() is called dev is NULL and so trigger the deref via
> > cmos->irq
> >
> > The problem comes from that the device is removed but no remove function
> > are called due to _exit_p().
> >
> > This patch remove all _exit_p() annotation.
> >
> > Signed-off-by: Corentin Labbe <[email protected]>
> > ---
>
>
> The changelog should go here.
>
> > drivers/rtc/rtc-cmos.c | 10 +++++-----
> > 1 file changed, 5 insertions(+), 5 deletions(-)
> >
Sorry, could I just resent it with PATCH RESENT v2 ? or does I need to bump it to v3.
Regards
On 18/10/2016 at 16:24:53 +0200, LABBE Corentin wrote :
> On Tue, Oct 18, 2016 at 04:16:39PM +0200, Alexandre Belloni wrote:
> > Hi,
> >
> > On 18/10/2016 at 13:55:53 +0200, Corentin Labbe wrote :
> > > Change since v1:
> > > - restored __exit for module_exit function
> > >
> > > ---8<---
> >
> > I don't think this is the proper scissor symbol for git. At least, it is
> > no recognized by patchwork and messes your patch.
> >
> > > From c04a24acebee4b4432afa7a2a7d4cfe114cc9a40 Mon Sep 17 00:00:00 2001
> > > From: Corentin Labbe <[email protected]>
> > > Date: Mon, 17 Oct 2016 14:17:49 +0200
> > > Subject: [PATCH v2] rtc: cmos: remove all __exit_p annotations
> > >
> > > I got the following stack trace under qemu:
> >
> > [...]
> > > When cmos_set_alarm() is called dev is NULL and so trigger the deref via
> > > cmos->irq
> > >
> > > The problem comes from that the device is removed but no remove function
> > > are called due to _exit_p().
> > >
> > > This patch remove all _exit_p() annotation.
> > >
> > > Signed-off-by: Corentin Labbe <[email protected]>
> > > ---
> >
> >
> > The changelog should go here.
> >
> > > drivers/rtc/rtc-cmos.c | 10 +++++-----
> > > 1 file changed, 5 insertions(+), 5 deletions(-)
> > >
>
> Sorry, could I just resent it with PATCH RESENT v2 ? or does I need to bump it to v3.
>
Resending v2 is fine
--
Alexandre Belloni, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com