* Re: [PATCH] loop: mask loop_control_ioctl parameter only as minor
2021-11-18 2:36 [PATCH] loop: mask loop_control_ioctl parameter only as minor wangyangbo
@ 2021-11-18 14:15 ` Tetsuo Handa
2021-11-18 14:51 ` Tetsuo Handa
2021-11-19 1:50 ` wangyangbo
2021-11-19 2:24 ` [PATCH] loop: check loop_control_ioctl parameter in range of minor wangyangbo
2021-11-19 2:32 ` [PATCH v2] " wangyangbo
2 siblings, 2 replies; 8+ messages in thread
From: Tetsuo Handa @ 2021-11-18 14:15 UTC (permalink / raw)
To: wangyangbo, axboe; +Cc: linux-block, linux-kernel
On 2021/11/18 11:36, wangyangbo wrote:
> @@ -2170,11 +2170,11 @@ static long loop_control_ioctl(struct file *file, unsigned int cmd,
> {
> switch (cmd) {
> case LOOP_CTL_ADD:
> - return loop_add(parm);
> + return loop_add(MINOR(parm));
Better to return -EINVAL or something if out of minor range?
> case LOOP_CTL_REMOVE:
> - return loop_control_remove(parm);
> + return loop_control_remove(MINOR(parm));
This is bad, for this change makes
if (idx < 0) {
pr_warn("deleting an unspecified loop device is not supported.\n");
return -EINVAL;
}
dead code by masking the argument to 0-1048575 range.
> case LOOP_CTL_GET_FREE:
> - return loop_control_get_free(parm);
> + return loop_control_get_free(MINOR(parm));
This is pointless, for the passed argument is not used.
By the way, didn't someone already propose removal of this argument?
> default:
> return -ENOSYS;
> }
>
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] loop: mask loop_control_ioctl parameter only as minor
2021-11-18 14:15 ` Tetsuo Handa
@ 2021-11-18 14:51 ` Tetsuo Handa
2021-11-18 15:21 ` wang yangbo
2021-11-19 7:15 ` Christoph Hellwig
2021-11-19 1:50 ` wangyangbo
1 sibling, 2 replies; 8+ messages in thread
From: Tetsuo Handa @ 2021-11-18 14:51 UTC (permalink / raw)
To: Luis Chamberlain; +Cc: linux-block, wangyangbo, axboe
On 2021/11/18 23:15, Tetsuo Handa wrote:
> On 2021/11/18 11:36, wangyangbo wrote:
>> @@ -2170,11 +2170,11 @@ static long loop_control_ioctl(struct file *file, unsigned int cmd,
>> {
>> switch (cmd) {
>> case LOOP_CTL_ADD:
>> - return loop_add(parm);
>> + return loop_add(MINOR(parm));
>
> Better to return -EINVAL or something if out of minor range?
Well, this is not specific to loop devices.
Shouldn't the minor range be checked by device_add_disk() ?
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] loop: mask loop_control_ioctl parameter only as minor
2021-11-18 14:51 ` Tetsuo Handa
@ 2021-11-18 15:21 ` wang yangbo
2021-11-19 7:15 ` Christoph Hellwig
1 sibling, 0 replies; 8+ messages in thread
From: wang yangbo @ 2021-11-18 15:21 UTC (permalink / raw)
To: Tetsuo Handa; +Cc: Luis Chamberlain, linux-block, axboe
> On Nov 18, 2021, at 22:51, Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp> wrote:
>
> On 2021/11/18 23:15, Tetsuo Handa wrote:
>>> On 2021/11/18 11:36, wangyangbo wrote:
>>> @@ -2170,11 +2170,11 @@ static long loop_control_ioctl(struct file *file, unsigned int cmd,
>>> {
>>> switch (cmd) {
>>> case LOOP_CTL_ADD:
>>> - return loop_add(parm);
>>> + return loop_add(MINOR(parm));
>>
>> Better to return -EINVAL or something if out of minor range?
>
> Well, this is not specific to loop devices.
> Shouldn't the minor range be checked by device_add_disk() ?
I just think ioctl paramter need to make sense.
Perhaps block layer need add/update/del check for consistency, but as inner interface caller check is also agreeable.
By the way, do other driver have similar problems?
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] loop: mask loop_control_ioctl parameter only as minor
2021-11-18 14:51 ` Tetsuo Handa
2021-11-18 15:21 ` wang yangbo
@ 2021-11-19 7:15 ` Christoph Hellwig
1 sibling, 0 replies; 8+ messages in thread
From: Christoph Hellwig @ 2021-11-19 7:15 UTC (permalink / raw)
To: Tetsuo Handa; +Cc: Luis Chamberlain, linux-block, wangyangbo, axboe
On Thu, Nov 18, 2021 at 11:51:21PM +0900, Tetsuo Handa wrote:
> On 2021/11/18 23:15, Tetsuo Handa wrote:
> > On 2021/11/18 11:36, wangyangbo wrote:
> >> @@ -2170,11 +2170,11 @@ static long loop_control_ioctl(struct file *file, unsigned int cmd,
> >> {
> >> switch (cmd) {
> >> case LOOP_CTL_ADD:
> >> - return loop_add(parm);
> >> + return loop_add(MINOR(parm));
> >
> > Better to return -EINVAL or something if out of minor range?
>
> Well, this is not specific to loop devices.
> Shouldn't the minor range be checked by device_add_disk() ?
Yes, that probably makes sense.
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: [PATCH] loop: mask loop_control_ioctl parameter only as minor
2021-11-18 14:15 ` Tetsuo Handa
2021-11-18 14:51 ` Tetsuo Handa
@ 2021-11-19 1:50 ` wangyangbo
1 sibling, 0 replies; 8+ messages in thread
From: wangyangbo @ 2021-11-19 1:50 UTC (permalink / raw)
To: Tetsuo Handa; +Cc: axboe, linux-block, linux-kernel
> On Nov 18, 2021, at 22:15, Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp> wrote:
>
> On 2021/11/18 11:36, wangyangbo wrote:
>> @@ -2170,11 +2170,11 @@ static long loop_control_ioctl(struct file *file, unsigned int cmd,
>> {
>> switch (cmd) {
>> case LOOP_CTL_ADD:
>> - return loop_add(parm);
>> + return loop_add(MINOR(parm));
>
> Better to return -EINVAL or something if out of minor range?
Definitely, EINVAL or EDOM, which do you think is better?
>
>> case LOOP_CTL_REMOVE:
>> - return loop_control_remove(parm);
>> + return loop_control_remove(MINOR(parm));
>
> This is bad, for this change makes
>
> if (idx < 0) {
> pr_warn("deleting an unspecified loop device is not supported.\n");
> return -EINVAL;
> }
>
> dead code by masking the argument to 0-1048575 range.
But ioctl param is unsigned long, I think this need to sanitize.
>> case LOOP_CTL_GET_FREE:
>> - return loop_control_get_free(parm);
>> + return loop_control_get_free(MINOR(parm));
>
> This is pointless, for the passed argument is not used.
> By the way, didn't someone already propose removal of this argument?
I don't find this in mail list, but I would like to sanitize that code.
^ permalink raw reply [flat|nested] 8+ messages in thread
* [PATCH] loop: check loop_control_ioctl parameter in range of minor
2021-11-18 2:36 [PATCH] loop: mask loop_control_ioctl parameter only as minor wangyangbo
2021-11-18 14:15 ` Tetsuo Handa
@ 2021-11-19 2:24 ` wangyangbo
2021-11-19 2:32 ` [PATCH v2] " wangyangbo
2 siblings, 0 replies; 8+ messages in thread
From: wangyangbo @ 2021-11-19 2:24 UTC (permalink / raw)
To: axboe; +Cc: penguin-kernel, linux-block, linux-kernel, wangyangbo
loop_control_ioctl call add_disk of block layer, but may pass number beyond MAX MINOR.
So check loop_control_ioctl parameter in range of minor, and delete redundant code.
Reproduce:
touch file
losetup /dev/loop1048576 file
losetup /dev/loop0 file
Problem:
sysfs: cannot create duplicate filename '/dev/block/7:0'
CPU: 0 PID: 529 Comm: losetup Not tainted 5.14.16-arch1-1 #1 ad87b876fa2ab6fdbd995dc1c9aab0ad8f767b2c
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
Call Trace:
dump_stack_lvl+0x46/0x5a
sysfs_warn_dup.cold+0x17/0x24
sysfs_do_create_link_sd+0xbe/0xd0
device_add+0x580/0x970
? dev_set_name+0x5b/0x80
__device_add_disk+0xb5/0x2f0
loop_add+0x236/0x290 [loop 2ed923fc8fbd84fdc093bf55ac085973636a3936]
loop_control_ioctl+0x7f/0x1f0 [loop 2ed923fc8fbd84fdc093bf55ac085973636a3936]
__x64_sys_ioctl+0x82/0xb0
do_syscall_64+0x5c/0x80
? syscall_exit_to_user_mode+0x23/0x40
? do_syscall_64+0x69/0x80
entry_SYSCALL_64_after_hwframe+0x44/0xae
RIP: 0033:0x7ff3e8ba259b
Code: ff ff ff 85 c0 79 9b 49 c7 c4 ff ff ff ff 5b 5d 4c 89 e0 41 5c c3 66 0f 1f 84 00 00 00 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d a5 a8 0c 00 f7 d8 64 89 01 48
RSP: 002b:00007ffe6d2eaa18 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 00007ffe6d2eb038 RCX: 00007ff3e8ba259b
RDX: 0000000000000000 RSI: 0000000000004c80 RDI: 0000000000000003
RBP: 00007ffe6d2eaae0 R08: 1999999999999999 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000003
R13: 00007ffe6d2eaa24 R14: 0000560b131bd200 R15: 0000560b131c44a0
---
drivers/block/loop.c | 12 +++++-------
1 file changed, 5 insertions(+), 7 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index a154cab6cd98..3b19ffaa63e9 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -2102,11 +2102,6 @@ static int loop_control_remove(int idx)
struct loop_device *lo;
int ret;
- if (idx < 0) {
- pr_warn("deleting an unspecified loop device is not supported.\n");
- return -EINVAL;
- }
-
/* Hide this loop device for serialization. */
ret = mutex_lock_killable(&loop_ctl_mutex);
if (ret)
@@ -2145,7 +2140,7 @@ static int loop_control_remove(int idx)
return ret;
}
-static int loop_control_get_free(int idx)
+static int loop_control_get_free(void)
{
struct loop_device *lo;
int id, ret;
@@ -2168,13 +2163,16 @@ static int loop_control_get_free(int idx)
static long loop_control_ioctl(struct file *file, unsigned int cmd,
unsigned long parm)
{
+ if (parm > MINORMASK)
+ pr_warn("ioctl parameter is out of max_minor.\n");
+ return -EINVAL;
switch (cmd) {
case LOOP_CTL_ADD:
return loop_add(parm);
case LOOP_CTL_REMOVE:
return loop_control_remove(parm);
case LOOP_CTL_GET_FREE:
- return loop_control_get_free(parm);
+ return loop_control_get_free();
default:
return -ENOSYS;
}
--
2.20.1
^ permalink raw reply related [flat|nested] 8+ messages in thread
* [PATCH v2] loop: check loop_control_ioctl parameter in range of minor
2021-11-18 2:36 [PATCH] loop: mask loop_control_ioctl parameter only as minor wangyangbo
2021-11-18 14:15 ` Tetsuo Handa
2021-11-19 2:24 ` [PATCH] loop: check loop_control_ioctl parameter in range of minor wangyangbo
@ 2021-11-19 2:32 ` wangyangbo
2 siblings, 0 replies; 8+ messages in thread
From: wangyangbo @ 2021-11-19 2:32 UTC (permalink / raw)
To: axboe; +Cc: penguin-kernel, linux-block, linux-kernel, wangyangbo
loop_control_ioctl call add_disk of block layer, but may pass number beyond MAX MINOR.
So check loop_control_ioctl parameter in range of minor, and delete redundant code.
Reproduce:
touch file
losetup /dev/loop1048576 file
losetup /dev/loop0 file
Problem:
sysfs: cannot create duplicate filename '/dev/block/7:0'
CPU: 0 PID: 529 Comm: losetup Not tainted 5.14.16-arch1-1 #1 ad87b876fa2ab6fdbd995dc1c9aab0ad8f767b2c
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
Call Trace:
dump_stack_lvl+0x46/0x5a
sysfs_warn_dup.cold+0x17/0x24
sysfs_do_create_link_sd+0xbe/0xd0
device_add+0x580/0x970
? dev_set_name+0x5b/0x80
__device_add_disk+0xb5/0x2f0
loop_add+0x236/0x290 [loop 2ed923fc8fbd84fdc093bf55ac085973636a3936]
loop_control_ioctl+0x7f/0x1f0 [loop 2ed923fc8fbd84fdc093bf55ac085973636a3936]
__x64_sys_ioctl+0x82/0xb0
do_syscall_64+0x5c/0x80
? syscall_exit_to_user_mode+0x23/0x40
? do_syscall_64+0x69/0x80
entry_SYSCALL_64_after_hwframe+0x44/0xae
RIP: 0033:0x7ff3e8ba259b
Code: ff ff ff 85 c0 79 9b 49 c7 c4 ff ff ff ff 5b 5d 4c 89 e0 41 5c c3 66 0f 1f 84 00 00 00 00 00 f3 0f 1e fa b8 10 00 00 00 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 8b 0d a5 a8 0c 00 f7 d8 64 89 01 48
RSP: 002b:00007ffe6d2eaa18 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
RAX: ffffffffffffffda RBX: 00007ffe6d2eb038 RCX: 00007ff3e8ba259b
RDX: 0000000000000000 RSI: 0000000000004c80 RDI: 0000000000000003
RBP: 00007ffe6d2eaae0 R08: 1999999999999999 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000003
R13: 00007ffe6d2eaa24 R14: 0000560b131bd200 R15: 0000560b131c44a0
---
drivers/block/loop.c | 13 ++++++-------
1 file changed, 6 insertions(+), 7 deletions(-)
diff --git a/drivers/block/loop.c b/drivers/block/loop.c
index a154cab6cd98..9e9d164b2a65 100644
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -2102,11 +2102,6 @@ static int loop_control_remove(int idx)
struct loop_device *lo;
int ret;
- if (idx < 0) {
- pr_warn("deleting an unspecified loop device is not supported.\n");
- return -EINVAL;
- }
-
/* Hide this loop device for serialization. */
ret = mutex_lock_killable(&loop_ctl_mutex);
if (ret)
@@ -2145,7 +2140,7 @@ static int loop_control_remove(int idx)
return ret;
}
-static int loop_control_get_free(int idx)
+static int loop_control_get_free(void)
{
struct loop_device *lo;
int id, ret;
@@ -2168,13 +2163,17 @@ static int loop_control_get_free(int idx)
static long loop_control_ioctl(struct file *file, unsigned int cmd,
unsigned long parm)
{
+ if (parm > MINORMASK) {
+ pr_warn("ioctl parameter is out of max_minor.\n");
+ return -EINVAL;
+ }
switch (cmd) {
case LOOP_CTL_ADD:
return loop_add(parm);
case LOOP_CTL_REMOVE:
return loop_control_remove(parm);
case LOOP_CTL_GET_FREE:
- return loop_control_get_free(parm);
+ return loop_control_get_free();
default:
return -ENOSYS;
}
--
2.20.1
^ permalink raw reply related [flat|nested] 8+ messages in thread