Re: [PATCH v3 03/10] crash_dump: Disallow writing to dm-crypt configfs during kexec_file_load syscall

[Coiby Xu <coiby.xu@gmail.com>]

On Wed, Aug 05, 2026 at 04:00:51PM +0530, Sourabh Jain wrote:
>
>
>On 29/07/26 09:06, Coiby Xu wrote:
>>If writing to the configfs group happens concurrently during
>>kexec_file_load syscall, it may lead to the following issues,
>>   - buffer overflow if dm-crypt keys are added after allocation
>>   - stale total_keys if dm-crypt keys are removed during iteration
>>   - keys_header will not be freed if config/crash_dm_crypt_key/reuse is
>>     set true
>>
>>So hold config_keys_subsys.su_mutex for the entire sequence during the
>>kexec_file_load syscall to ensure a consistent snapshot.
>>
>>Fixes: 479e58549b0f ("crash_dump: store dm crypt keys in kdump reserved memory")
>>Suggested-by: Sourabh Jain <sourabhjain@linux.ibm.com>
>>Signed-off-by: Coiby Xu <coiby.xu@gmail.com>
>>---
>>  kernel/crash_dump_dm_crypt.c | 23 +++++++++++++++++++++--
>>  1 file changed, 21 insertions(+), 2 deletions(-)
>>
>>diff --git a/kernel/crash_dump_dm_crypt.c b/kernel/crash_dump_dm_crypt.c
>>index 4335b6cb1fc4..d2e66c6fe6f3 100644
>>--- a/kernel/crash_dump_dm_crypt.c
>>+++ b/kernel/crash_dump_dm_crypt.c
>>@@ -293,6 +293,7 @@ static ssize_t config_keys_reuse_show(struct config_item *item, char *page)
>>  static ssize_t config_keys_reuse_store(struct config_item *item,
>>  					   const char *page, size_t count)
>>  {
>>+	struct mutex *lock;
>>  	bool val;
>>  	int r;
>>@@ -302,8 +303,12 @@ static ssize_t config_keys_reuse_store(struct config_item *item,
>>  		return -EINVAL;
>>  	}
>>+	lock = &to_config_group(item)->cg_subsys->su_mutex;
>>+	mutex_lock(lock);
>
>Is this lock only protecting against races between key reuse and 
>kexec_file_load(),

The lock here is to protect against races between key reuse and
kexec_file_load.

>or does it also handle the case where a new key is added during key reuse or
>kexec_file_load() is running?

For the cases where a key is added/deleted, configfs will automatically
take care of them because it will acquire mutex lock automatically.

>
>If it is only intended to protect key reuse versus kexec_file_load(), 
>why can't we use
>the kexec lock instead?
>
>The reason I'm asking is that, in upcoming patches, the key reuse path 
>accesses
>kexec_crash_image properties and the crash reserved region directly. 
>Doing so
>without taking the kexec lock (using kexec_trylock()) could lead to 
>race conditions.

After comparing the kexec lock approach with the configfs mutex lock
approach, I think the latter is a simpler solution because
1. the kexec lock is non-blocking and we have to repeatedly try until the
    lock get acquired. So it means user space has to make changes as
    well.

2. configfs already acquires the mutex lock automatically for
    creating/deleting configfs items. So if we use configfs mutex lock,
    it means one less place to use the lock.

In config_keys_reuse_store, kexec_crash_image will be checked before
accessing its properties and the crash reserved region. Can you
elaborate on what the race conditions are? Will acquiring the lock
before accessing kexec_crash_image properties and the crash reserved
region help protect against these races?

In theory, the kexec lock can be a more robust approach. But considering
only root can write to the crash dm-crypt keys configfs and load kdump
image, I'm not sure it's necessary to adopt a bit more complex solution.


>
>- Sourabh Jain
>>+
>>+	r = -EINVAL;
>>  	if (kstrtobool(page, &val) || !val)
>>-		return -EINVAL;
>>+		goto unlock;
>
>The jump above skips setting count, causing the function to return 
>count instead of -EINVAL.
>Is this really intended?

Thanks for catching this issue! In the end of the function, r instead of
count should be returned.

>
>>  	if (is_dm_key_reused) {
>>  		pr_info("Already got dm-crypt keys, please continue with kexec_file_load syscall\n");
>>@@ -311,11 +316,15 @@ static ssize_t config_keys_reuse_store(struct config_item *item,
>>  		r = get_keys_from_kdump_reserved_memory();
>>  		if (r) {
>>  			pr_warn("Failed to get dm-crypt keys from reserved memory\n");
>>-			return r;
>>+			goto unlock;
>>  		}
>>  		is_dm_key_reused = true;
>>  	}
>>+	r = count;
>>+
>>+unlock:
>>+	mutex_unlock(lock);
>>  	return count;
>>  }
>>@@ -421,6 +430,8 @@ static int build_keys_header(void)
>>  	return 0;
>>  }
>>+static bool mutex_acquired;
>>+
>>  int crash_load_dm_crypt_keys(struct kimage *image)
>>  {
>>  	struct kexec_buf kbuf = {
>>@@ -432,6 +443,9 @@ int crash_load_dm_crypt_keys(struct kimage *image)
>>  	};
>>  	int r = 0;
>>+	mutex_lock(&config_keys_subsys.su_mutex);
>>+	mutex_acquired = true;
>>+
>>  	if (key_count <= 0) {
>>  		kexec_dprintk("No dm-crypt keys\n");
>>  		return 0;
>>@@ -481,6 +495,11 @@ void kexec_file_post_load_cleanup_dm_crypt(struct kimage *image)
>>  		kfree_sensitive(keys_header);
>>  		keys_header = NULL;
>>  	}
>>+
>>+	if (mutex_acquired) {
>>+		mutex_unlock(&config_keys_subsys.su_mutex);
>>+		mutex_acquired = false;
>>+	}
>>  }
>>  static int __init configfs_dmcrypt_keys_init(void)
>

-- 
Best regards,
Coiby