[PATCH 2/2] debugfs: Remove broken no-mount mode

Aaron Thompson posted 2 patches 1 week, 4 days ago
[PATCH 2/2] debugfs: Remove broken no-mount mode
Posted by Aaron Thompson 1 week, 4 days ago
debugfs access modes were added in Linux 5.10 (Dec 2020) [1], but the
no-mount mode has behaved effectively the same as the off mode since
Linux 5.12 (Apr 2021) [2]. The only difference is the specific error
code returned by the debugfs_create_* functions, which is -ENOENT in
no-mount mode and -EPERM in off mode.

Given that no-mount hasn't worked for several years with no complaints,
just remove it.

[1] a24c6f7bc923 ("debugfs: Add access restriction option")

[2] bc6de804d36b ("debugfs: be more robust at handling improper input in debugfs_lookup()")
    56348560d495 ("debugfs: do not attempt to create a new file before the filesystem is initalized")

Signed-off-by: Aaron Thompson <dev@aaront.org>
---
 .../admin-guide/kernel-parameters.txt          |  6 +-----
 fs/debugfs/inode.c                             | 18 +++++++++++-------
 fs/debugfs/internal.h                          | 13 -------------
 lib/Kconfig.debug                              |  9 +--------
 4 files changed, 13 insertions(+), 33 deletions(-)

diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
index 6c42061ca20e..847a17efe289 100644
--- a/Documentation/admin-guide/kernel-parameters.txt
+++ b/Documentation/admin-guide/kernel-parameters.txt
@@ -1113,12 +1113,8 @@
 
 	debugfs=    	[KNL,EARLY] This parameter enables what is exposed to
 			userspace and debugfs internal clients.
-			Format: { on, no-mount, off }
+			Format: { on, off }
 			on: 	All functions are enabled.
-			no-mount:
-				Filesystem is not registered but kernel clients can
-			        access APIs and a crashkernel can be used to read
-				its content. There is nothing to mount.
 			off: 	Filesystem is not registered and clients
 			        get a -EPERM as result when trying to register files
 				or directories within debugfs.
diff --git a/fs/debugfs/inode.c b/fs/debugfs/inode.c
index b6e401c46b6b..0284b0256195 100644
--- a/fs/debugfs/inode.c
+++ b/fs/debugfs/inode.c
@@ -35,7 +35,7 @@
 static struct vfsmount *debugfs_mount;
 static int debugfs_mount_count;
 static bool debugfs_registered;
-static unsigned int debugfs_allow __ro_after_init = DEFAULT_DEBUGFS_ALLOW_BITS;
+static bool debugfs_enabled __ro_after_init = IS_ENABLED(DEBUG_FS_ALLOW_ALL);
 
 /*
  * Don't allow access attributes to be changed whilst the kernel is locked down
@@ -365,7 +365,7 @@ static struct dentry *debugfs_start_creating(const char *name,
 	struct dentry *dentry;
 	int error;
 
-	if (!(debugfs_allow & DEBUGFS_ALLOW_API))
+	if (!debugfs_enabled)
 		return ERR_PTR(-EPERM);
 
 	if (!debugfs_initialized())
@@ -885,21 +885,25 @@ static int __init debugfs_kernel(char *str)
 {
 	if (str) {
 		if (!strcmp(str, "on"))
-			debugfs_allow = DEBUGFS_ALLOW_API | DEBUGFS_ALLOW_MOUNT;
-		else if (!strcmp(str, "no-mount"))
-			debugfs_allow = DEBUGFS_ALLOW_API;
+			debugfs_enabled = true;
 		else if (!strcmp(str, "off"))
-			debugfs_allow = 0;
+			debugfs_enabled = false;
+		else if (!strcmp(str, "no-mount")) {
+			pr_notice("debugfs=no-mount is a deprecated alias "
+				  "for debugfs=off\n");
+			debugfs_enabled = false;
+		}
 	}
 
 	return 0;
 }
 early_param("debugfs", debugfs_kernel);
+
 static int __init debugfs_init(void)
 {
 	int retval;
 
-	if (!(debugfs_allow & DEBUGFS_ALLOW_MOUNT))
+	if (!debugfs_enabled)
 		return -EPERM;
 
 	retval = sysfs_create_mount_point(kernel_kobj, "debug");
diff --git a/fs/debugfs/internal.h b/fs/debugfs/internal.h
index 427987f81571..c95699b27a56 100644
--- a/fs/debugfs/internal.h
+++ b/fs/debugfs/internal.h
@@ -55,17 +55,4 @@ enum {
 	HAS_IOCTL = 16
 };
 
-#define DEBUGFS_ALLOW_API	BIT(0)
-#define DEBUGFS_ALLOW_MOUNT	BIT(1)
-
-#ifdef CONFIG_DEBUG_FS_ALLOW_ALL
-#define DEFAULT_DEBUGFS_ALLOW_BITS (DEBUGFS_ALLOW_MOUNT | DEBUGFS_ALLOW_API)
-#endif
-#ifdef CONFIG_DEBUG_FS_DISALLOW_MOUNT
-#define DEFAULT_DEBUGFS_ALLOW_BITS (DEBUGFS_ALLOW_API)
-#endif
-#ifdef CONFIG_DEBUG_FS_ALLOW_NONE
-#define DEFAULT_DEBUGFS_ALLOW_BITS (0)
-#endif
-
 #endif /* _DEBUGFS_INTERNAL_H_ */
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 3034e294d50d..d9ab42916143 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -679,7 +679,7 @@ choice
 	help
 	  This selects the default access restrictions for debugfs.
 	  It can be overridden with kernel command line option
-	  debugfs=[on,no-mount,off]. The restrictions apply for API access
+	  debugfs=[on,off]. The restrictions apply for API access
 	  and filesystem registration.
 
 config DEBUG_FS_ALLOW_ALL
@@ -688,13 +688,6 @@ config DEBUG_FS_ALLOW_ALL
 	  No restrictions apply. Both API and filesystem registration
 	  is on. This is the normal default operation.
 
-config DEBUG_FS_DISALLOW_MOUNT
-	bool "Do not register debugfs as filesystem"
-	help
-	  The API is open but filesystem is not loaded. Clients can still do
-	  their work and read with debug tools that do not need
-	  debugfs filesystem.
-
 config DEBUG_FS_ALLOW_NONE
 	bool "No access"
 	help
-- 
2.47.3
Re: [PATCH 2/2] debugfs: Remove broken no-mount mode
Posted by Mark Brown 8 hours ago
On Thu, Nov 20, 2025 at 10:26:33AM +0000, Aaron Thompson wrote:

> debugfs access modes were added in Linux 5.10 (Dec 2020) [1], but the
> no-mount mode has behaved effectively the same as the off mode since
> Linux 5.12 (Apr 2021) [2]. The only difference is the specific error
> code returned by the debugfs_create_* functions, which is -ENOENT in
> no-mount mode and -EPERM in off mode.

I'm seeing regressions in -next in a lot of testing stuff which bisect
to this patch.  I've got a test that looks at the deferred probe list to
see if it's empty, and the mm split_huge_page_test which uses a debugfs
file called split_huge_pages.  Neither of these mount debugfs for
themselves, they just assume it'll be there - it looks like that's not
happening any more but I didn't investigate properly.

I don't immediately see what's getting confused, DEBUG_FS_ALLOW_ALL is
the default and not overridden by anything in any defconfig so
debugfs_enabled still ought to be being set, but I didn't actually try
to debug this yet.

Sample bisect:

git bisect start
# status: waiting for both good and bad commits
# bad: [95cb2fd6ce0ad61af54191fe5ef271d7177f9c3a] Add linux-next specific files for 20251201
git bisect bad 95cb2fd6ce0ad61af54191fe5ef271d7177f9c3a
# status: waiting for good commit(s), bad commit known
# good: [4ffc97b7687d1b355f08f464e212fb1085ec5c34] Merge branch 'tip/urgent' of https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git
git bisect good 4ffc97b7687d1b355f08f464e212fb1085ec5c34
# good: [87d5c4addc7e535618586e7205191a7f402288ba] Merge branch 'master' of https://git.kernel.org/pub/scm/linux/kernel/git/herbert/cryptodev-2.6.git
git bisect good 87d5c4addc7e535618586e7205191a7f402288ba
# good: [a4ad48eac682ccdc21e2f16b8f27abbf615d8d3d] Merge branch 'for-next' of https://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator.git
git bisect good a4ad48eac682ccdc21e2f16b8f27abbf615d8d3d
# bad: [b99f4ac0a6c7ccf37be14f5ef61b160b1c8a74b0] Merge branch 'driver-core-next' of https://git.kernel.org/pub/scm/linux/kernel/git/driver-core/driver-core.git
git bisect bad b99f4ac0a6c7ccf37be14f5ef61b160b1c8a74b0
# good: [24cefd05bbf969c95fff3733da174e8a352c1cb2] Merge branch 'master' of https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git
git bisect good 24cefd05bbf969c95fff3733da174e8a352c1cb2
# good: [a5d0f36158717334aa85753fab8dedfa0dfeeaca] Merge branch 'riscv_kvm_next' of https://github.com/kvm-riscv/linux.git
git bisect good a5d0f36158717334aa85753fab8dedfa0dfeeaca
# good: [f16919001053bba074af696571831d4d3e7f86bc] Merge branch 'for-next' of https://git.kernel.org/pub/scm/linux/kernel/git/tj/wq.git
git bisect good f16919001053bba074af696571831d4d3e7f86bc
# good: [4f7961a0cc56637fb5df8d118e72ab12160e6c11] Merge branch 'for-next' of https://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86.git
git bisect good 4f7961a0cc56637fb5df8d118e72ab12160e6c11
# good: [0a75f3d90e7ab9cd182327fca4b4e3bce379afe5] devres: Move devm_alloc_percpu() and related to devres.h
git bisect good 0a75f3d90e7ab9cd182327fca4b4e3bce379afe5
# good: [2a522e38f813c7fab9e49ebdfbfaba4c05b06579] Merge branch 'for-leds-next' of https://git.kernel.org/pub/scm/linux/kernel/git/lee/leds.git
git bisect good 2a522e38f813c7fab9e49ebdfbfaba4c05b06579
# good: [f10c23fa159c5481dfe0025e619dc5ef844f6ce1] tick/nohz: avoid showing '(null)' if nohz_full= not set
git bisect good f10c23fa159c5481dfe0025e619dc5ef844f6ce1
# good: [f940e425c6fbcf5282e8daddb9351239e2343598] Merge branch 'for-next' of https://github.com/cminyard/linux-ipmi.git
git bisect good f940e425c6fbcf5282e8daddb9351239e2343598
# good: [ac1ab906d7a98e34be95ef63b81ff828cc432346] driver core: WQ_PERCPU added to alloc_workqueue users
git bisect good ac1ab906d7a98e34be95ef63b81ff828cc432346
# good: [3ae94a55d047d133fad1e6c811befe4347b75791] debugfs: Remove redundant access mode checks
git bisect good 3ae94a55d047d133fad1e6c811befe4347b75791
# bad: [f278809475f6835b56de78b28dc2cc0c7e2c20a4] debugfs: Remove broken no-mount mode
git bisect bad f278809475f6835b56de78b28dc2cc0c7e2c20a4
# first bad commit: [f278809475f6835b56de78b28dc2cc0c7e2c20a4] debugfs: Remove broken no-mount mode
Re: [PATCH 2/2] debugfs: Remove broken no-mount mode
Posted by Aaron Thompson 3 hours ago
On 12/1/25 09:15, Mark Brown wrote:
> On Thu, Nov 20, 2025 at 10:26:33AM +0000, Aaron Thompson wrote:
> 
>> debugfs access modes were added in Linux 5.10 (Dec 2020) [1], but the
>> no-mount mode has behaved effectively the same as the off mode since
>> Linux 5.12 (Apr 2021) [2]. The only difference is the specific error
>> code returned by the debugfs_create_* functions, which is -ENOENT in
>> no-mount mode and -EPERM in off mode.
> 
> I'm seeing regressions in -next in a lot of testing stuff which bisect
> to this patch.  I've got a test that looks at the deferred probe list to
> see if it's empty, and the mm split_huge_page_test which uses a debugfs
> file called split_huge_pages.  Neither of these mount debugfs for
> themselves, they just assume it'll be there - it looks like that's not
> happening any more but I didn't investigate properly.
> 
> I don't immediately see what's getting confused, DEBUG_FS_ALLOW_ALL is
> the default and not overridden by anything in any defconfig so
> debugfs_enabled still ought to be being set, but I didn't actually try
> to debug this yet.
> 
> Sample bisect:
> 
> git bisect start
> # status: waiting for both good and bad commits
> ...
> # bad: [f278809475f6835b56de78b28dc2cc0c7e2c20a4] debugfs: Remove broken no-mount mode
> git bisect bad f278809475f6835b56de78b28dc2cc0c7e2c20a4
> # first bad commit: [f278809475f6835b56de78b28dc2cc0c7e2c20a4] debugfs: Remove broken no-mount mode

I am terribly sorry, this was a sloppy mistake on my part. The 
IS_ENABLED() check is missing the CONFIG_ prefix. The fix patch is attached.

Greg, should I send a v2 of the patch series, or a separate patch with 
just the fix? Or something else? Again, sorry for the trouble.

-- Aaron