Since debugfs include sensitive information it need to be treated
carefully. But it also has many very useful debug functions for userspace.
With this option we can have same configuration for system with
need of debugfs and a way to turn it off. It is needed new
kernel command line parameter to be activated.
Signed-off-by: Peter Enderborg <[email protected]>
---
fs/debugfs/inode.c | 17 ++++++++++++++++-
lib/Kconfig.debug | 10 ++++++++++
2 files changed, 26 insertions(+), 1 deletion(-)
diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
index b7f2e971ecbc..bde37dab77e0 100644
--- a/fs/debugfs/inode.c
+++ b/fs/debugfs/inode.c
@@ -786,10 +786,25 @@ bool debugfs_initialized(void)
}
EXPORT_SYMBOL_GPL(debugfs_initialized);
+static int allow_debugfs;
+
+static int __init debugfs_kernel(char *str)
+{
+ if (str && !strcmp(str, "true"))
+ allow_debugfs = true;
+
+ return 0;
+
+}
+early_param("debugfs", debugfs_kernel);
+
static int __init debugfs_init(void)
{
int retval;
-
+#ifdef CONFIG_DEBUG_FS_MOUNT_RESTRICTED
+ if (!allow_debugfs)
+ return -EPERM;
+#endif
retval = sysfs_create_mount_point(kernel_kobj, "debug");
if (retval)
return retval;
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 21d9c5f6e7ec..d3a3338740d2 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -443,6 +443,16 @@ config DEBUG_FS
If unsure, say N.
+config DEBUG_FS_MOUNT_RESTRICTED
+ bool "Debug Filesystem mount restricted"
+ depends on DEBUG_FS
+ help
+ This is an additional restriction for mounting debugfs. It allows
+ the kernel to have debugfs compiled, but requires that kernel command
+ line has a debugfs parameter to register as a filesystem.
+
+ If unsure, say N.
+
source "lib/Kconfig.kgdb"
source "lib/Kconfig.ubsan"
--
2.26.2
On Thu, May 28, 2020 at 10:00:31AM +0200, Peter Enderborg wrote:
> Since debugfs include sensitive information it need to be treated
> carefully. But it also has many very useful debug functions for userspace.
> With this option we can have same configuration for system with
> need of debugfs and a way to turn it off. It is needed new
> kernel command line parameter to be activated.
By "configuration" do you mean "kernel configuration"? What is wrong
with relying on the build option like we do today?
You might want to reword all of this to make more sense about the
"problem" you are trying to solve here, as I don't really understand it,
sorry.
>
> Signed-off-by: Peter Enderborg <[email protected]>
> ---
> fs/debugfs/inode.c | 17 ++++++++++++++++-
> lib/Kconfig.debug | 10 ++++++++++
> 2 files changed, 26 insertions(+), 1 deletion(-)
No documentation update? That's not good :(
> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
> index b7f2e971ecbc..bde37dab77e0 100644
> --- a/fs/debugfs/inode.c
> +++ b/fs/debugfs/inode.c
> @@ -786,10 +786,25 @@ bool debugfs_initialized(void)
> }
> EXPORT_SYMBOL_GPL(debugfs_initialized);
>
> +static int allow_debugfs;
> +
> +static int __init debugfs_kernel(char *str)
> +{
> + if (str && !strcmp(str, "true"))
> + allow_debugfs = true;
> +
> + return 0;
> +
> +}
> +early_param("debugfs", debugfs_kernel);
> +
> static int __init debugfs_init(void)
> {
> int retval;
> -
> +#ifdef CONFIG_DEBUG_FS_MOUNT_RESTRICTED
> + if (!allow_debugfs)
> + return -EPERM;
> +#endif
But you are not restricting the ability to mount it here, you are
removing the ability for it to even start up at all. What does this
break for code that thinks the filesystem is registered (i.e. the call
to simple_pin_fs() in start_creating() in fs/debugfs/inode.c?
> retval = sysfs_create_mount_point(kernel_kobj, "debug");
> if (retval)
> return retval;
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 21d9c5f6e7ec..d3a3338740d2 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -443,6 +443,16 @@ config DEBUG_FS
>
> If unsure, say N.
>
> +config DEBUG_FS_MOUNT_RESTRICTED
> + bool "Debug Filesystem mount restricted"
> + depends on DEBUG_FS
> + help
> + This is an additional restriction for mounting debugfs. It allows
> + the kernel to have debugfs compiled, but requires that kernel command
> + line has a debugfs parameter to register as a filesystem.
> +
> + If unsure, say N.
No hint as to what the command line option is? And again, it's not a
restriction for mounting it, you are preventing it from initializing
entirely.
Which, as I have found out over time, can had side affects for codepaths
that try to check the return value of debugfs calls, and if they fail to
succeed, abort the whole driver/subsystem/whatever initializaion path,
which is not what I think you want to do here at all.
thanks,
greg k-h
On 5/28/20 10:27 AM, Greg Kroah-Hartman wrote:
> On Thu, May 28, 2020 at 10:00:31AM +0200, Peter Enderborg wrote:
>> Since debugfs include sensitive information it need to be treated
>> carefully. But it also has many very useful debug functions for userspace.
>> With this option we can have same configuration for system with
>> need of debugfs and a way to turn it off. It is needed new
>> kernel command line parameter to be activated.
> By "configuration" do you mean "kernel configuration"? What is wrong
> with relying on the build option like we do today?
>
> You might want to reword all of this to make more sense about the
> "problem" you are trying to solve here, as I don't really understand it,
> sorry.
>
>
>> Signed-off-by: Peter Enderborg <[email protected]>
>> ---
>> fs/debugfs/inode.c | 17 ++++++++++++++++-
>> lib/Kconfig.debug | 10 ++++++++++
>> 2 files changed, 26 insertions(+), 1 deletion(-)
> No documentation update? That's not good :(
>
>
>> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
>> index b7f2e971ecbc..bde37dab77e0 100644
>> --- a/fs/debugfs/inode.c
>> +++ b/fs/debugfs/inode.c
>> @@ -786,10 +786,25 @@ bool debugfs_initialized(void)
>> }
>> EXPORT_SYMBOL_GPL(debugfs_initialized);
>>
>> +static int allow_debugfs;
>> +
>> +static int __init debugfs_kernel(char *str)
>> +{
>> + if (str && !strcmp(str, "true"))
>> + allow_debugfs = true;
>> +
>> + return 0;
>> +
>> +}
>> +early_param("debugfs", debugfs_kernel);
>> +
>> static int __init debugfs_init(void)
>> {
>> int retval;
>> -
>> +#ifdef CONFIG_DEBUG_FS_MOUNT_RESTRICTED
>> + if (!allow_debugfs)
>> + return -EPERM;
>> +#endif
> But you are not restricting the ability to mount it here, you are
> removing the ability for it to even start up at all. What does this
> break for code that thinks the filesystem is registered (i.e. the call
> to simple_pin_fs() in start_creating() in fs/debugfs/inode.c?
>
If it does, the lines below is also cause the same problem.
>
>> retval = sysfs_create_mount_point(kernel_kobj, "debug");
>> if (retval)
>> return retval;
>> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
>> index 21d9c5f6e7ec..d3a3338740d2 100644
>> --- a/lib/Kconfig.debug
>> +++ b/lib/Kconfig.debug
>> @@ -443,6 +443,16 @@ config DEBUG_FS
>>
>> If unsure, say N.
>>
>> +config DEBUG_FS_MOUNT_RESTRICTED
>> + bool "Debug Filesystem mount restricted"
>> + depends on DEBUG_FS
>> + help
>> + This is an additional restriction for mounting debugfs. It allows
>> + the kernel to have debugfs compiled, but requires that kernel command
>> + line has a debugfs parameter to register as a filesystem.
>> +
>> + If unsure, say N.
> No hint as to what the command line option is? And again, it's not a
> restriction for mounting it, you are preventing it from initializing
> entirely.
>
> Which, as I have found out over time, can had side affects for codepaths
> that try to check the return value of debugfs calls, and if they fail to
> succeed, abort the whole driver/subsystem/whatever initializaion path,
> which is not what I think you want to do here at all.
>
> thanks,
>
> greg k-h
On Thu, May 28, 2020 at 08:39:02AM +0000, Enderborg, Peter wrote:
> On 5/28/20 10:27 AM, Greg Kroah-Hartman wrote:
> > On Thu, May 28, 2020 at 10:00:31AM +0200, Peter Enderborg wrote:
> >> Since debugfs include sensitive information it need to be treated
> >> carefully. But it also has many very useful debug functions for userspace.
> >> With this option we can have same configuration for system with
> >> need of debugfs and a way to turn it off. It is needed new
> >> kernel command line parameter to be activated.
> > By "configuration" do you mean "kernel configuration"? What is wrong
> > with relying on the build option like we do today?
> >
> > You might want to reword all of this to make more sense about the
> > "problem" you are trying to solve here, as I don't really understand it,
> > sorry.
> >
> >
> >> Signed-off-by: Peter Enderborg <[email protected]>
> >> ---
> >> fs/debugfs/inode.c | 17 ++++++++++++++++-
> >> lib/Kconfig.debug | 10 ++++++++++
> >> 2 files changed, 26 insertions(+), 1 deletion(-)
> > No documentation update? That's not good :(
> >
> >
> >> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
> >> index b7f2e971ecbc..bde37dab77e0 100644
> >> --- a/fs/debugfs/inode.c
> >> +++ b/fs/debugfs/inode.c
> >> @@ -786,10 +786,25 @@ bool debugfs_initialized(void)
> >> }
> >> EXPORT_SYMBOL_GPL(debugfs_initialized);
> >>
> >> +static int allow_debugfs;
> >> +
> >> +static int __init debugfs_kernel(char *str)
> >> +{
> >> + if (str && !strcmp(str, "true"))
> >> + allow_debugfs = true;
> >> +
> >> + return 0;
> >> +
> >> +}
> >> +early_param("debugfs", debugfs_kernel);
> >> +
> >> static int __init debugfs_init(void)
> >> {
> >> int retval;
> >> -
> >> +#ifdef CONFIG_DEBUG_FS_MOUNT_RESTRICTED
> >> + if (!allow_debugfs)
> >> + return -EPERM;
> >> +#endif
> > But you are not restricting the ability to mount it here, you are
> > removing the ability for it to even start up at all. What does this
> > break for code that thinks the filesystem is registered (i.e. the call
> > to simple_pin_fs() in start_creating() in fs/debugfs/inode.c?
> >
> If it does, the lines below is also cause the same problem.
In a working system, errors in the lines below will never happen :)
thanks,
greg k-h
On 5/28/20 11:27 AM, Greg Kroah-Hartman wrote:
> On Thu, May 28, 2020 at 08:39:02AM +0000, Enderborg, Peter wrote:
>> On 5/28/20 10:27 AM, Greg Kroah-Hartman wrote:
>>> On Thu, May 28, 2020 at 10:00:31AM +0200, Peter Enderborg wrote:
>>>> Since debugfs include sensitive information it need to be treated
>>>> carefully. But it also has many very useful debug functions for userspace.
>>>> With this option we can have same configuration for system with
>>>> need of debugfs and a way to turn it off. It is needed new
>>>> kernel command line parameter to be activated.
>>> By "configuration" do you mean "kernel configuration"? What is wrong
>>> with relying on the build option like we do today?
>>>
>>> You might want to reword all of this to make more sense about the
>>> "problem" you are trying to solve here, as I don't really understand it,
>>> sorry.
>>>
>>>
>>>> Signed-off-by: Peter Enderborg <[email protected]>
>>>> ---
>>>> fs/debugfs/inode.c | 17 ++++++++++++++++-
>>>> lib/Kconfig.debug | 10 ++++++++++
>>>> 2 files changed, 26 insertions(+), 1 deletion(-)
>>> No documentation update? That's not good :(
>>>
>>>
>>>> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
>>>> index b7f2e971ecbc..bde37dab77e0 100644
>>>> --- a/fs/debugfs/inode.c
>>>> +++ b/fs/debugfs/inode.c
>>>> @@ -786,10 +786,25 @@ bool debugfs_initialized(void)
>>>> }
>>>> EXPORT_SYMBOL_GPL(debugfs_initialized);
>>>>
>>>> +static int allow_debugfs;
>>>> +
>>>> +static int __init debugfs_kernel(char *str)
>>>> +{
>>>> + if (str && !strcmp(str, "true"))
>>>> + allow_debugfs = true;
>>>> +
>>>> + return 0;
>>>> +
>>>> +}
>>>> +early_param("debugfs", debugfs_kernel);
>>>> +
>>>> static int __init debugfs_init(void)
>>>> {
>>>> int retval;
>>>> -
>>>> +#ifdef CONFIG_DEBUG_FS_MOUNT_RESTRICTED
>>>> + if (!allow_debugfs)
>>>> + return -EPERM;
>>>> +#endif
>>> But you are not restricting the ability to mount it here, you are
>>> removing the ability for it to even start up at all. What does this
>>> break for code that thinks the filesystem is registered (i.e. the call
>>> to simple_pin_fs() in start_creating() in fs/debugfs/inode.c?
>>>
>> If it does, the lines below is also cause the same problem.
> In a working system, errors in the lines below will never happen :)
I think it is not unreasonable to expect that debug functionality are robust enough to handle
system where not everything is working smoothly.
> thanks,
>
> greg k-h
On Thu, May 28, 2020 at 12:04:32PM +0000, Enderborg, Peter wrote:
> On 5/28/20 11:27 AM, Greg Kroah-Hartman wrote:
> > On Thu, May 28, 2020 at 08:39:02AM +0000, Enderborg, Peter wrote:
> >> On 5/28/20 10:27 AM, Greg Kroah-Hartman wrote:
> >>> On Thu, May 28, 2020 at 10:00:31AM +0200, Peter Enderborg wrote:
> >>>> Since debugfs include sensitive information it need to be treated
> >>>> carefully. But it also has many very useful debug functions for userspace.
> >>>> With this option we can have same configuration for system with
> >>>> need of debugfs and a way to turn it off. It is needed new
> >>>> kernel command line parameter to be activated.
> >>> By "configuration" do you mean "kernel configuration"? What is wrong
> >>> with relying on the build option like we do today?
> >>>
> >>> You might want to reword all of this to make more sense about the
> >>> "problem" you are trying to solve here, as I don't really understand it,
> >>> sorry.
> >>>
> >>>
> >>>> Signed-off-by: Peter Enderborg <[email protected]>
> >>>> ---
> >>>> fs/debugfs/inode.c | 17 ++++++++++++++++-
> >>>> lib/Kconfig.debug | 10 ++++++++++
> >>>> 2 files changed, 26 insertions(+), 1 deletion(-)
> >>> No documentation update? That's not good :(
> >>>
> >>>
> >>>> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
> >>>> index b7f2e971ecbc..bde37dab77e0 100644
> >>>> --- a/fs/debugfs/inode.c
> >>>> +++ b/fs/debugfs/inode.c
> >>>> @@ -786,10 +786,25 @@ bool debugfs_initialized(void)
> >>>> }
> >>>> EXPORT_SYMBOL_GPL(debugfs_initialized);
> >>>>
> >>>> +static int allow_debugfs;
> >>>> +
> >>>> +static int __init debugfs_kernel(char *str)
> >>>> +{
> >>>> + if (str && !strcmp(str, "true"))
> >>>> + allow_debugfs = true;
> >>>> +
> >>>> + return 0;
> >>>> +
> >>>> +}
> >>>> +early_param("debugfs", debugfs_kernel);
> >>>> +
> >>>> static int __init debugfs_init(void)
> >>>> {
> >>>> int retval;
> >>>> -
> >>>> +#ifdef CONFIG_DEBUG_FS_MOUNT_RESTRICTED
> >>>> + if (!allow_debugfs)
> >>>> + return -EPERM;
> >>>> +#endif
> >>> But you are not restricting the ability to mount it here, you are
> >>> removing the ability for it to even start up at all. What does this
> >>> break for code that thinks the filesystem is registered (i.e. the call
> >>> to simple_pin_fs() in start_creating() in fs/debugfs/inode.c?
> >>>
> >> If it does, the lines below is also cause the same problem.
> > In a working system, errors in the lines below will never happen :)
>
> I think it is not unreasonable to expect that debug functionality are robust enough to handle
> system where not everything is working smoothly.
I do not disagree, but have you tested this to ensure that your kernel
still works when you disable debugfs in this manner?
If so, great! Just trying to point out a potential problem...
thanks,
greg k-h
On 5/28/20 2:23 PM, Greg Kroah-Hartman wrote:
> On Thu, May 28, 2020 at 12:04:32PM +0000, Enderborg, Peter wrote:
>> On 5/28/20 11:27 AM, Greg Kroah-Hartman wrote:
>>> On Thu, May 28, 2020 at 08:39:02AM +0000, Enderborg, Peter wrote:
>>>> On 5/28/20 10:27 AM, Greg Kroah-Hartman wrote:
>>>>> On Thu, May 28, 2020 at 10:00:31AM +0200, Peter Enderborg wrote:
>>>>>> Since debugfs include sensitive information it need to be treated
>>>>>> carefully. But it also has many very useful debug functions for userspace.
>>>>>> With this option we can have same configuration for system with
>>>>>> need of debugfs and a way to turn it off. It is needed new
>>>>>> kernel command line parameter to be activated.
>>>>> By "configuration" do you mean "kernel configuration"? What is wrong
>>>>> with relying on the build option like we do today?
>>>>>
>>>>> You might want to reword all of this to make more sense about the
>>>>> "problem" you are trying to solve here, as I don't really understand it,
>>>>> sorry.
>>>>>
>>>>>
>>>>>> Signed-off-by: Peter Enderborg <[email protected]>
>>>>>> ---
>>>>>> fs/debugfs/inode.c | 17 ++++++++++++++++-
>>>>>> lib/Kconfig.debug | 10 ++++++++++
>>>>>> 2 files changed, 26 insertions(+), 1 deletion(-)
>>>>> No documentation update? That's not good :(
>>>>>
>>>>>
>>>>>> diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
>>>>>> index b7f2e971ecbc..bde37dab77e0 100644
>>>>>> --- a/fs/debugfs/inode.c
>>>>>> +++ b/fs/debugfs/inode.c
>>>>>> @@ -786,10 +786,25 @@ bool debugfs_initialized(void)
>>>>>> }
>>>>>> EXPORT_SYMBOL_GPL(debugfs_initialized);
>>>>>>
>>>>>> +static int allow_debugfs;
>>>>>> +
>>>>>> +static int __init debugfs_kernel(char *str)
>>>>>> +{
>>>>>> + if (str && !strcmp(str, "true"))
>>>>>> + allow_debugfs = true;
>>>>>> +
>>>>>> + return 0;
>>>>>> +
>>>>>> +}
>>>>>> +early_param("debugfs", debugfs_kernel);
>>>>>> +
>>>>>> static int __init debugfs_init(void)
>>>>>> {
>>>>>> int retval;
>>>>>> -
>>>>>> +#ifdef CONFIG_DEBUG_FS_MOUNT_RESTRICTED
>>>>>> + if (!allow_debugfs)
>>>>>> + return -EPERM;
>>>>>> +#endif
>>>>> But you are not restricting the ability to mount it here, you are
>>>>> removing the ability for it to even start up at all. What does this
>>>>> break for code that thinks the filesystem is registered (i.e. the call
>>>>> to simple_pin_fs() in start_creating() in fs/debugfs/inode.c?
>>>>>
>>>> If it does, the lines below is also cause the same problem.
>>> In a working system, errors in the lines below will never happen :)
>> I think it is not unreasonable to expect that debug functionality are robust enough to handle
>> system where not everything is working smoothly.
> I do not disagree, but have you tested this to ensure that your kernel
> still works when you disable debugfs in this manner?
So far only tested on fedora f32 (x86_64) and older arm64 systems.
So far only zswap that prints out that "debugfs initialization failed" and it still works fine.
> If so, great! Just trying to point out a potential problem...
Will do new patchset, im waiting a bit longer to see if there some other comments first.
> thanks,
>
> greg k-h