fs/read_all: Filter /dev/watchdog*

Message ID 1521025644-18312-1-git-send-email-xuyang.jy@cn.fujitsu.com
State Superseded
Headers show
Series
  • fs/read_all: Filter /dev/watchdog*
Related show

Commit Message

yang xu March 14, 2018, 11:07 a.m.
On some distros with Magic Close feature or built-in CONFIG_WATCHDOG_NOWAYOUT,
just closing /dev/watchdog* enabled by open leads to system reboot as expected.

If Magic Close feature is supported, just writing a specific magic character 'V'
into /dev/watchdog* before closing it can disable the watchdog.

If CONFIG_WATCHDOG_NOWAYOUT is built-in, there is no way to disable the watchdog.

Magic Close feature is introduced by:
commit 017cf080("watchDog Timer Driver Core - Add Magic Close feature")

Please see the following url for detailed watchdog info:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/watchdog/watchdog-api.txt

Signed-off-by: yang xu <xuyang.jy@cn.fujitsu.com>
Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
---
 testcases/kernel/fs/read_all/read_all.c | 3 +++
 1 file changed, 3 insertions(+)

Comments

Richard Palethorpe March 15, 2018, 12:25 p.m. | #1
Hello,

yang xu writes:

> On some distros with Magic Close feature or built-in CONFIG_WATCHDOG_NOWAYOUT,
> just closing /dev/watchdog* enabled by open leads to system reboot as expected.
>
> If Magic Close feature is supported, just writing a specific magic character 'V'
> into /dev/watchdog* before closing it can disable the watchdog.
>
> If CONFIG_WATCHDOG_NOWAYOUT is built-in, there is no way to disable the watchdog.
>
> Magic Close feature is introduced by:
> commit 017cf080("watchDog Timer Driver Core - Add Magic Close feature")
>
> Please see the following url for detailed watchdog info:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/watchdog/watchdog-api.txt
>
> Signed-off-by: yang xu <xuyang.jy@cn.fujitsu.com>
> Signed-off-by: Xiao Yang <yangx.jy@cn.fujitsu.com>
> ---
>  testcases/kernel/fs/read_all/read_all.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/testcases/kernel/fs/read_all/read_all.c b/testcases/kernel/fs/read_all/read_all.c
> index 81806e7..a841b88 100644
> --- a/testcases/kernel/fs/read_all/read_all.c
> +++ b/testcases/kernel/fs/read_all/read_all.c
> @@ -393,6 +393,9 @@ static void visit_dir(const char *path)
>  		snprintf(dent_path, MAX_PATH,
>  			 "%s/%s", path, dent->d_name);
>
> +		if (!strncmp(dent_path, "/dev/watchdog", 13))
> +			continue;
> +

I don't think this should be hardcoded because it is OK to read this
file on some systems. Unfortunately there does not seem to be a reliable
way to find out if it is OK (/sys/class/watchdog/watchdogn/nowayout is
not even present on my system)

However perhaps we could set the default for the exclude parameter (-e)
to /dev/watchdog? Then the user can easily override it, but we won't
reboot some people's systems by default. What do you think?

>  		if (act == DA_UNKNOWN) {
>  			if (lstat(dent_path, &dent_st))
>  				tst_res(TINFO | TERRNO, "lstat(%s)", path);
> --
> 1.8.3.1


--
Thank you,
Richard.
yang xu March 16, 2018, 5:18 a.m. | #2
2018/3/15 20:25, Richard Palethorpe write:
> Hello,
>
> yang xu writes:
>
>> On some distros with Magic Close feature or built-in CONFIG_WATCHDOG_NOWAYOUT,
>> just closing /dev/watchdog* enabled by open leads to system reboot as expected.
>>
>> If Magic Close feature is supported, just writing a specific magic character 'V'
>> into /dev/watchdog* before closing it can disable the watchdog.
>>
>> If CONFIG_WATCHDOG_NOWAYOUT is built-in, there is no way to disable the watchdog.
>>
>> Magic Close feature is introduced by:
>> commit 017cf080("watchDog Timer Driver Core - Add Magic Close feature")
>>
>> Please see the following url for detailed watchdog info:
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/watchdog/watchdog-api.txt
>>
>> Signed-off-by: yang xu<xuyang.jy@cn.fujitsu.com>
>> Signed-off-by: Xiao Yang<yangx.jy@cn.fujitsu.com>
>> ---
>>   testcases/kernel/fs/read_all/read_all.c | 3 +++
>>   1 file changed, 3 insertions(+)
>>
>> diff --git a/testcases/kernel/fs/read_all/read_all.c b/testcases/kernel/fs/read_all/read_all.c
>> index 81806e7..a841b88 100644
>> --- a/testcases/kernel/fs/read_all/read_all.c
>> +++ b/testcases/kernel/fs/read_all/read_all.c
>> @@ -393,6 +393,9 @@ static void visit_dir(const char *path)
>>   		snprintf(dent_path, MAX_PATH,
>>   			 "%s/%s", path, dent->d_name);
>>
>> +		if (!strncmp(dent_path, "/dev/watchdog", 13))
>> +			continue;
>> +
> I don't think this should be hardcoded because it is OK to read this
> file on some systems. Unfortunately there does not seem to be a reliable
> way to find out if it is OK (/sys/class/watchdog/watchdogn/nowayout is
> not even present on my system)
>
> However perhaps we could set the default for the exclude parameter (-e)
> to /dev/watchdog? Then the user can easily override it, but we won't
> reboot some people's systems by default. What do you think?

Hi Richard
Agreed. We can run read_all with the -e option to exclude the 
/dev/watchdog*, as bleow:

diff --git a/runtest/fs b/runtest/fs
index a595edb..42a9bfc 100644
--- a/runtest/fs
+++ b/runtest/fs
@@ -69,7 +69,7 @@ fs_di fs_di -d $TMPDIR
# Was not sure why it should reside in runtest/crashme and won´t get 
tested ever
proc01 proc01 -m 128

-read_all_dev read_all -d /dev -q -r 10
+read_all_dev read_all -d /dev -e '/dev/watchdog?(0)' -q -r 10
read_all_proc read_all -d /proc -q -r 10
read_all_sys read_all -d /sys -q -r 10

How about this modification?

Best Regards,
Yang Xu

>>   		if (act == DA_UNKNOWN) {
>>   			if (lstat(dent_path,&dent_st))
>>   				tst_res(TINFO | TERRNO, "lstat(%s)", path);
>> --
>> 1.8.3.1
>
> --
> Thank you,
> Richard.
>
>
> .
>
Richard Palethorpe March 19, 2018, 9:14 a.m. | #3
Hello,

xuyang.jy writes:

> 2018/3/15 20:25, Richard Palethorpe write:
>> Hello,
>>
>> yang xu writes:
>>
>>> On some distros with Magic Close feature or built-in CONFIG_WATCHDOG_NOWAYOUT,
>>> just closing /dev/watchdog* enabled by open leads to system reboot as expected.
>>>
>>> If Magic Close feature is supported, just writing a specific magic character 'V'
>>> into /dev/watchdog* before closing it can disable the watchdog.
>>>
>>> If CONFIG_WATCHDOG_NOWAYOUT is built-in, there is no way to disable the watchdog.
>>>
>>> Magic Close feature is introduced by:
>>> commit 017cf080("watchDog Timer Driver Core - Add Magic Close feature")
>>>
>>> Please see the following url for detailed watchdog info:
>>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/watchdog/watchdog-api.txt
>>>
>>> Signed-off-by: yang xu<xuyang.jy@cn.fujitsu.com>
>>> Signed-off-by: Xiao Yang<yangx.jy@cn.fujitsu.com>
>>> ---
>>>   testcases/kernel/fs/read_all/read_all.c | 3 +++
>>>   1 file changed, 3 insertions(+)
>>>
>>> diff --git a/testcases/kernel/fs/read_all/read_all.c b/testcases/kernel/fs/read_all/read_all.c
>>> index 81806e7..a841b88 100644
>>> --- a/testcases/kernel/fs/read_all/read_all.c
>>> +++ b/testcases/kernel/fs/read_all/read_all.c
>>> @@ -393,6 +393,9 @@ static void visit_dir(const char *path)
>>>   		snprintf(dent_path, MAX_PATH,
>>>   			 "%s/%s", path, dent->d_name);
>>>
>>> +		if (!strncmp(dent_path, "/dev/watchdog", 13))
>>> +			continue;
>>> +
>> I don't think this should be hardcoded because it is OK to read this
>> file on some systems. Unfortunately there does not seem to be a reliable
>> way to find out if it is OK (/sys/class/watchdog/watchdogn/nowayout is
>> not even present on my system)
>>
>> However perhaps we could set the default for the exclude parameter (-e)
>> to /dev/watchdog? Then the user can easily override it, but we won't
>> reboot some people's systems by default. What do you think?
>
> Hi Richard
> Agreed. We can run read_all with the -e option to exclude the /dev/watchdog*,
> as bleow:
>
> diff --git a/runtest/fs b/runtest/fs
> index a595edb..42a9bfc 100644
> --- a/runtest/fs
> +++ b/runtest/fs
> @@ -69,7 +69,7 @@ fs_di fs_di -d $TMPDIR
> # Was not sure why it should reside in runtest/crashme and won´t get tested
> ever
> proc01 proc01 -m 128
>
> -read_all_dev read_all -d /dev -q -r 10
> +read_all_dev read_all -d /dev -e '/dev/watchdog?(0)' -q -r 10
> read_all_proc read_all -d /proc -q -r 10
> read_all_sys read_all -d /sys -q -r 10
>
> How about this modification?

I think that is OK, but Cyril would prefer to make it the default value
of the exclude variable so that it does not cause a restart when the
user runs read_all manually.

I think I still prefer putting it in the runtest file because the exception
is only relevant to /dev and I don't want to have exceptions hardcoded.

>
> Best Regards,
> Yang Xu
>
>>>   		if (act == DA_UNKNOWN) {
>>>   			if (lstat(dent_path,&dent_st))
>>>   				tst_res(TINFO | TERRNO, "lstat(%s)", path);
>>> --
>>> 1.8.3.1
>>
>> --
>> Thank you,
>> Richard.
>>
>>
>> .
>>

Patch

diff --git a/testcases/kernel/fs/read_all/read_all.c b/testcases/kernel/fs/read_all/read_all.c
index 81806e7..a841b88 100644
--- a/testcases/kernel/fs/read_all/read_all.c
+++ b/testcases/kernel/fs/read_all/read_all.c
@@ -393,6 +393,9 @@  static void visit_dir(const char *path)
 		snprintf(dent_path, MAX_PATH,
 			 "%s/%s", path, dent->d_name);
 
+		if (!strncmp(dent_path, "/dev/watchdog", 13))
+			continue;
+
 		if (act == DA_UNKNOWN) {
 			if (lstat(dent_path, &dent_st))
 				tst_res(TINFO | TERRNO, "lstat(%s)", path);