On 1/24/22 14:21, Denis Pauk wrote:
On Sun, 23 Jan 2022 18:32:48 -0800
Guenter Roeck <linux@xxxxxxxxxxxx> wrote:
Paweł Marciniak reports the following crash, observed when clearing
the chassis intrusion alarm.
BUG: kernel NULL pointer dereference, address: 0000000000000028
PGD 0 P4D 0
Oops: 0000 [#1] PREEMPT SMP PTI
CPU: 3 PID: 4815 Comm: bash Tainted: G S
5.16.2-200.fc35.x86_64 #1 Hardware name: To Be Filled By O.E.M. To Be
Filled By O.E.M./Z97 Extreme4, BIOS P2.60A 05/03/2018 RIP:
0010:clear_caseopen+0x5a/0x120 [nct6775] Code: 68 70 e8 e9 32 b1 e3
85 c0 0f 85 d2 00 00 00 48 83 7c 24 ... RSP: 0018:ffffabcb02803dd8
EFLAGS: 00010246 RAX: 0000000000000000 RBX: 0000000000000002 RCX:
0000000000000000 RDX: ffff8e8808192880 RSI: 0000000000000000 RDI:
ffff8e87c7509a68 RBP: 0000000000000000 R08: 0000000000000001 R09:
000000000000000a R10: 000000000000000a R11: f000000000000000 R12:
000000000000001f R13: ffff8e87c7509828 R14: ffff8e87c7509a68 R15:
ffff8e88494527a0 FS: 00007f4db9151740(0000)
GS:ffff8e8ebfec0000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000
ES: 0000 CR0: 0000000080050033 CR2: 0000000000000028 CR3:
0000000166b66001 CR4: 00000000001706e0 Call Trace:
<TASK>
kernfs_fop_write_iter+0x11c/0x1b0
new_sync_write+0x10b/0x180
vfs_write+0x209/0x2a0
ksys_write+0x4f/0xc0
do_syscall_64+0x3b/0x90
entry_SYSCALL_64_after_hwframe+0x44/0xae
The problem is that the device passed to clear_caseopen() is the hwmon
device, not the platform device, and the platform data is not set in
the hwmon device. Store the pointer to sio_data in struct
nct6775_data and get if from there if needed.
Fixes: 2e7b9886968b ("hwmon: (nct6775) Use superio_*() function
pointers in sio_data.") Cc: Denis Pauk <pauk.denis@xxxxxxxxx>
Cc: Bernhard Seibold <mail@xxxxxxxxxxxxxxxxxxx>
Reported-by: Paweł Marciniak <pmarciniak@xxxxxxxxxxxx>
Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
---
drivers/hwmon/nct6775.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/hwmon/nct6775.c b/drivers/hwmon/nct6775.c
index fd3f91cb01c6..098d12b9ecda 100644
--- a/drivers/hwmon/nct6775.c
+++ b/drivers/hwmon/nct6775.c
@@ -1175,7 +1175,7 @@ static inline u8 in_to_reg(u32 val, u8 nr)
struct nct6775_data {
int addr; /* IO base of hw monitor block */
- int sioreg; /* SIO register address */
+ struct nct6775_sio_data *sio_data;
enum kinds kind;
What do you think about remove kind field also from struct
nct6775_data? Code has copy of this field inside of the struct
nct6775_sio_data.
I thought about it, but it would add a level of indirection to
each access, and it is used for each register access (through
is_word_sized). That would make the change pretty expensive
in terms of runtime cost.
Guenter
const char *name;
@@ -3559,7 +3559,7 @@ clear_caseopen(struct device *dev, struct
device_attribute *attr, const char *buf, size_t count)
{
struct nct6775_data *data = dev_get_drvdata(dev);
- struct nct6775_sio_data *sio_data = dev_get_platdata(dev);
+ struct nct6775_sio_data *sio_data = data->sio_data;
int nr = to_sensor_dev_attr(attr)->index -
INTRUSION_ALARM_BASE; unsigned long val;
u8 reg;
@@ -3967,7 +3967,7 @@ static int nct6775_probe(struct platform_device
*pdev) return -ENOMEM;
data->kind = sio_data->kind;
- data->sioreg = sio_data->sioreg;
+ data->sio_data = sio_data;
if (sio_data->access == access_direct) {
data->addr = res->start;
Tested-by: Denis Pauk <pauk.denis@xxxxxxxxx>
Best regards,
Denis.
