Skip to contentSkip to conversation

ACPI: mrrm: Fix memory leaks and improve error handling

PATCH Superseded Kaushlendra Kumar 3 participants 7 messages Oct 7, 2025 – Oct 28, 2025 ACPI

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ACPI: mrrm: Fix memory leaks and improve error handling

Add proper error handling and resource cleanup to prevent memory leaks in add_boot_memory_ranges(). The function now checks for NULL return from kobject_create_and_add(), frees allocated names after use, and implements a cleanup path that removes previously created sysfs groups and kobjects on failure.

This prevents resource leaks when kobject creation or sysfs group creation fails during boot memory range initialization.

---
 drivers/acpi/acpi_mrrm.c | 33 +++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)
drivers/acpi/acpi_mrrm.c +29 -4
index 47ea3ccc2142..6ec42eb48783 100644
@@ -152,23 +152,48 @@ static __init int add_boot_memory_ranges(void)
 	struct kobject *pkobj, *kobj;
 	int ret = -EINVAL;
 	char *name;
+	int i;
 
 	pkobj = kobject_create_and_add("memory_ranges", acpi_kobj);
+	if (!pkobj)
+		return -ENOMEM;
 
-	for (int i = 0; i < mrrm_mem_entry_num; i++) {
+	for (i = 0; i < mrrm_mem_entry_num; i++) {
 		name = kasprintf(GFP_KERNEL, "range%d", i);
 		if (!name) {
 			ret = -ENOMEM;
-			break;
+			goto cleanup;
 		}
 
 		kobj = kobject_create_and_add(name, pkobj);
+		kfree(name);
+		if (!kobj) {
+			ret = -ENOMEM;
+			goto cleanup;
+		}
 
 		ret = sysfs_create_groups(kobj, memory_range_groups);
-		if (ret)
-			return ret;
+		if (ret) {
+			kobject_put(kobj);
+			goto cleanup;
+		}
 	}
 
+	return 0;
+
+cleanup:
+	for (int j = 0; j < i; j++) {
+		char cleanup_name[32];
+		struct kobject *cleanup_kobj;
+
+		snprintf(cleanup_name, sizeof(cleanup_name), "range%d", j);
+		cleanup_kobj = kobject_get(pkobj);
+		if (cleanup_kobj) {
+			sysfs_remove_groups(cleanup_kobj, memory_range_groups);
+			kobject_put(cleanup_kobj);
+		}
+	}
+	kobject_put(pkobj);
 	return ret;
 }
 
Signature · 1 line
2.34.1
#1 lore raw
Quoted 20 lines · Kaushlendra Kumar

Add proper error handling and resource cleanup to prevent memory leaks in add_boot_memory_ranges(). The function now checks for NULL return from kobject_create_and_add(), frees allocated names after use, and implements a cleanup path that removes previously created sysfs groups and kobjects on failure.

This prevents resource leaks when kobject creation or sysfs group creation fails during boot memory range initialization.

---
 drivers/acpi/acpi_mrrm.c | 33 +++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)
drivers/acpi/acpi_mrrm.c +0 -0
index 47ea3ccc2142..6ec42eb48783 100644

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 158

Quoted 10 lines
drivers/acpi/acpi_mrrm.c +3 -1
@@ -152,23 +152,48 @@ static __init int add_boot_memory_ranges(void)
        struct kobject *pkobj, *kobj;
        int ret = -EINVAL;
        char *name;
+       int i;
 
        pkobj = kobject_create_and_add("memory_ranges", acpi_kobj);
+       if (!pkobj)
+               return -ENOMEM;
 
-       for (int i = 0; i < mrrm_mem_entry_num; i++) {

Yes, i should be declared in the preamble.

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 169

Quoted 10 lines
drivers/acpi/acpi_mrrm.c +3 -1
+       for (i = 0; i < mrrm_mem_entry_num; i++) {
                name = kasprintf(GFP_KERNEL, "range%d", i);
                if (!name) {
                        ret = -ENOMEM;
-                       break;
+                       goto cleanup;
                }
 
                kobj = kobject_create_and_add(name, pkobj);
+               kfree(name);

OK, this fixes a memory leak.

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 172

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (!kobj) {
+                       ret = -ENOMEM;
+                       goto cleanup;

Why terminate this? Why not continue?

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 169

Quoted 5 lines
drivers/acpi/acpi_mrrm.c +1 -2
+               }
 
                ret = sysfs_create_groups(kobj, memory_range_groups);
-               if (ret)
-                       return ret;

Well, this returns already, but I'm not sure why. Tony, wouldn't it be better to continue?

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 178

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (ret) {
+                       kobject_put(kobj);
+                       goto cleanup;

I would do a "continue" instead.

Quoted 22 lines
drivers/acpi/acpi_mrrm.c +16 -1
+               }
        }
 
+       return 0;
+
+cleanup:
+       for (int j = 0; j < i; j++) {
+               char cleanup_name[32];
+               struct kobject *cleanup_kobj;
+
+               snprintf(cleanup_name, sizeof(cleanup_name), "range%d", j);
+               cleanup_kobj = kobject_get(pkobj);
+               if (cleanup_kobj) {
+                       sysfs_remove_groups(cleanup_kobj, memory_range_groups);
+                       kobject_put(cleanup_kobj);
+               }
+       }
+       kobject_put(pkobj);
        return ret;
 }
 
--
#2 lore raw
Quoted 43 lines · Rafael J. Wysocki
Quoted 20 lines · Kaushlendra Kumar

Add proper error handling and resource cleanup to prevent memory leaks in add_boot_memory_ranges(). The function now checks for NULL return from kobject_create_and_add(), frees allocated names after use, and implements a cleanup path that removes previously created sysfs groups and kobjects on failure.

This prevents resource leaks when kobject creation or sysfs group creation fails during boot memory range initialization.

---
 drivers/acpi/acpi_mrrm.c | 33 +++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)
drivers/acpi/acpi_mrrm.c +0 -0
index 47ea3ccc2142..6ec42eb48783 100644

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 158

Quoted 10 lines
drivers/acpi/acpi_mrrm.c +3 -1
@@ -152,23 +152,48 @@ static __init int add_boot_memory_ranges(void)
        struct kobject *pkobj, *kobj;
        int ret = -EINVAL;
        char *name;
+       int i;
 
        pkobj = kobject_create_and_add("memory_ranges", acpi_kobj);
+       if (!pkobj)
+               return -ENOMEM;
 
-       for (int i = 0; i < mrrm_mem_entry_num; i++) {

Yes, i should be declared in the preamble.

Quoted 8 lines
drivers/acpi/acpi_mrrm.c +2 -1
+       for (i = 0; i < mrrm_mem_entry_num; i++) {
                name = kasprintf(GFP_KERNEL, "range%d", i);
                if (!name) {
                        ret = -ENOMEM;
-                       break;
+                       goto cleanup;
                }
 
Quoted 4 lines

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 169

Quoted 2 lines
drivers/acpi/acpi_mrrm.c +1 -0
                kobj = kobject_create_and_add(name, pkobj);
+               kfree(name);

OK, this fixes a memory leak.

I didn't realize that kobject_create_and_add() made its own copy of the "name" parameter. Maybe the code should avoid the alloc/free by making "name" a local variable?

	char name[16] = "rangeXXXXXXXXXX";

	sprintf(&name[5], "%d", i);
Quoted 5 lines

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 172

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (!kobj) {
+                       ret = -ENOMEM;
+                       goto cleanup;

Why terminate this? Why not continue?

Terminating here (and below) will go to Kaushlendra's cleanup code to remove all the ranges. Maybe this is better than having some random subset of files appear (based on which allocations succeeded/failed)?

Quoted 36 lines

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 169

Quoted 5 lines
drivers/acpi/acpi_mrrm.c +1 -2
+               }
 
                ret = sysfs_create_groups(kobj, memory_range_groups);
-               if (ret)
-                       return ret;

Well, this returns already, but I'm not sure why. Tony, wouldn't it be better to continue?

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 178

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (ret) {
+                       kobject_put(kobj);
+                       goto cleanup;

I would do a "continue" instead.

Quoted 21 lines
drivers/acpi/acpi_mrrm.c +16 -0
+               }
        }
 
+       return 0;
+
+cleanup:
+       for (int j = 0; j < i; j++) {
+               char cleanup_name[32];
+               struct kobject *cleanup_kobj;
+
+               snprintf(cleanup_name, sizeof(cleanup_name), "range%d", j);
+               cleanup_kobj = kobject_get(pkobj);
+               if (cleanup_kobj) {
+                       sysfs_remove_groups(cleanup_kobj, memory_range_groups);
+                       kobject_put(cleanup_kobj);
+               }
+       }
+       kobject_put(pkobj);
        return ret;
 }
 
Quoted 1 line
Quoted 1 line

--

-Tony

#3 lore raw
Quoted 50 lines · Tony
Quoted 43 lines · Rafael J. Wysocki
Quoted 20 lines · Kaushlendra Kumar

Add proper error handling and resource cleanup to prevent memory leaks in add_boot_memory_ranges(). The function now checks for NULL return from kobject_create_and_add(), frees allocated names after use, and implements a cleanup path that removes previously created sysfs groups and kobjects on failure.

This prevents resource leaks when kobject creation or sysfs group creation fails during boot memory range initialization.

---
 drivers/acpi/acpi_mrrm.c | 33 +++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)
drivers/acpi/acpi_mrrm.c +0 -0
index 47ea3ccc2142..6ec42eb48783 100644

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 158

Quoted 10 lines
drivers/acpi/acpi_mrrm.c +3 -1
@@ -152,23 +152,48 @@ static __init int add_boot_memory_ranges(void)
        struct kobject *pkobj, *kobj;
        int ret = -EINVAL;
        char *name;
+       int i;
 
        pkobj = kobject_create_and_add("memory_ranges", acpi_kobj);
+       if (!pkobj)
+               return -ENOMEM;
 
-       for (int i = 0; i < mrrm_mem_entry_num; i++) {

Yes, i should be declared in the preamble.

Quoted 8 lines
drivers/acpi/acpi_mrrm.c +2 -1
+       for (i = 0; i < mrrm_mem_entry_num; i++) {
                name = kasprintf(GFP_KERNEL, "range%d", i);
                if (!name) {
                        ret = -ENOMEM;
-                       break;
+                       goto cleanup;
                }
 
Quoted 4 lines

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 169

Quoted 2 lines
drivers/acpi/acpi_mrrm.c +1 -0
                kobj = kobject_create_and_add(name, pkobj);
+               kfree(name);

OK, this fixes a memory leak.

Quoted 7 lines

I didn't realize that kobject_create_and_add() made its own copy of the "name" parameter. Maybe the code should avoid the alloc/free by making "name" a local variable?

        char name[16] = "rangeXXXXXXXXXX";

        sprintf(&name[5], "%d", i);

That'd work.

Quoted 10 lines
Quoted 5 lines

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 172

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (!kobj) {
+                       ret = -ENOMEM;
+                       goto cleanup;

Why terminate this? Why not continue?

Terminating here (and below) will go to Kaushlendra's cleanup code to remove all the ranges. Maybe this is better than having some random subset of files appear (based on which allocations succeeded/failed)?

Well, I guess it can be argued both ways and either one is fine with me. If that's what you prefer, let it be done, but the changelog needs a bit of work because the cleanup is done in error cases.

Quoted 39 lines
Quoted 36 lines

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 169

Quoted 5 lines
drivers/acpi/acpi_mrrm.c +1 -2
+               }
 
                ret = sysfs_create_groups(kobj, memory_range_groups);
-               if (ret)
-                       return ret;

Well, this returns already, but I'm not sure why. Tony, wouldn't it be better to continue?

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 178

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (ret) {
+                       kobject_put(kobj);
+                       goto cleanup;

I would do a "continue" instead.

Quoted 21 lines
drivers/acpi/acpi_mrrm.c +16 -0
+               }
        }
 
+       return 0;
+
+cleanup:
+       for (int j = 0; j < i; j++) {
+               char cleanup_name[32];
+               struct kobject *cleanup_kobj;
+
+               snprintf(cleanup_name, sizeof(cleanup_name), "range%d", j);
+               cleanup_kobj = kobject_get(pkobj);
+               if (cleanup_kobj) {
+                       sysfs_remove_groups(cleanup_kobj, memory_range_groups);
+                       kobject_put(cleanup_kobj);
+               }
+       }
+       kobject_put(pkobj);
        return ret;
 }
 
Quoted 1 line
Quoted 1 line

--

-Tony

#4 lore raw
In reply to Luck, Tony
Quoted 6 lines · Rafael J. Wysocki

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 172

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (!kobj) {
+                       ret = -ENOMEM;
+                       goto cleanup;

Why terminate this? Why not continue?

Quoted 7 lines

I didn't realize that kobject_create_and_add() made its own copy of the "name" parameter. Maybe the code should avoid the alloc/free by making "name" a local variable?

	char name[16] = "rangeXXXXXXXXXX";

	sprintf(&name[5], "%d", i);

Terminating here (and below) will go to Kaushlendra's cleanup code to remove all the ranges. Maybe this is better than having some random subset of files appear (based on which allocations succeeded/failed)?

Tony is exactly right here. I believe the current approach is more appropriate Here.

Regarding Tony's suggestion about the local buffer - that's a valid optimization, but I'd prefer to keep the current approach for clarity. The kasprintf/kfree pattern is more explicit about memory management in kernel.

The cleanup path ensures we maintain system consistency - either we successfully create the complete interface, or we leave the system in a clean state with no partial artifacts.

Thanks for the detailed review!

Kaushlendra

#5 lore raw
Quoted 13 lines
Quoted 11 lines · Rafael J. Wysocki

drivers/acpi/acpi_mrrm.c static __init int add_boot_memory_ranges(void) · line 172

Quoted 3 lines
drivers/acpi/acpi_mrrm.c +3 -0
+               if (!kobj) {
+                       ret = -ENOMEM;
+                       goto cleanup;

Why terminate this? Why not continue?

I didn't realize that kobject_create_and_add() made its own copy of the "name" parameter. Maybe the code should avoid the alloc/free by making "name" a local variable?

    char name[16] = "rangeXXXXXXXXXX";

    sprintf(&name[5], "%d", i);
Quoted 8 lines
Quoted 1 line

Terminating here (and below) will go to Kaushlendra's cleanup code to remove all the ranges. Maybe this is better than having some random subset of files appear (based on which allocations succeeded/failed)?

Tony is exactly right here. I believe the current approach is more appropriate Here.

Regarding Tony's suggestion about the local buffer - that's a valid optimization, but I'd prefer to keep the current approach for clarity. The kasprintf/kfree pattern is more explicit about memory management in kernel.

Why? I'll note that the cleanup code you added uses a local buffer "char cleanup_name[32];" to hold filenames (though now that I look back at the patch, that string isn't actually used!)

Quoted 6 lines

The cleanup path ensures we maintain system consistency - either we successfully create the complete interface, or we leave the system in a clean state with no partial artifacts.

Thanks for the detailed review!

-Tony

#6 lore raw
Quoted 2 lines · Tony Luck

Why? I'll note that the cleanup code you added uses a local buffer "char cleanup_name[32];" to hold filenames (though now that I look back at the patch, that string isn't actually used!)

Got it. Looking at the cleanup code again,
I see the inconsistency - I'm using a local buffer there but kasprintf() in the main loop.

I will use a local buffer approach.

Also thanks for catching the unused cleanup_name variable - I'll fix that in the cleanup logic as well.

I'll send a v2 with these improvements.

Thanks for the review!