kernel.git/fs/pstore/inode.c, branch linux-6.5.y

pstore: Add priv field to pstore_record for backend specific use

2022-06-24T18:40:04Z

The EFI pstore backend will need to store per-record variable name data when we switch away from the efivars layer. Add a priv field to struct pstore_record, and document it as holding a backend specific pointer that is assumed to be a kmalloc()d buffer, and will be kfree()d when the entire record is freed. Acked-by: Kees Cook Signed-off-by: Ard Biesheuvel

pstore: Fix warning in pstore_kill_sb()

2021-02-23T17:27:20Z

syzbot is hitting WARN_ON(pstore_sb != sb) at pstore_kill_sb() [1], for the assumption that pstore_sb != NULL is wrong because pstore_fill_super() will not assign pstore_sb = sb when new_inode() for d_make_root() returned NULL (due to memory allocation fault injection). Since mount_single() calls pstore_kill_sb() when pstore_fill_super() failed, pstore_kill_sb() needs to be aware of such failure path. [1] https://syzkaller.appspot.com/bug?id=6abacb8da5137cb47a416f2bef95719ed60508a0 Reported-by: syzbot Signed-off-by: Tetsuo Handa Signed-off-by: Kees Cook Link: https://lore.kernel.org/r/20210214031307.57903-1-penguin-kernel@I-love.SAKURA.ne.jp

pstore: Move kmsg_bytes default into Kconfig

2020-12-01T20:09:17Z

While kmsg_bytes can be set for pstore via mount, if a crash occurs before the mount only partial console log will be stored as kmsg_bytes defaults to a potentially different hardcoded value in the kernel (PSTORE_DEFAULT_KMSG_BYTES). This makes it impossible to analyze valuable post-mortem data especially on the embedded development or in the process of bringing up new boards. Change this value to be a Kconfig option with the default of old PSTORE_DEFAULT_KMSG_BYTES Signed-off-by: Vasile-Laurentiu Stanimir Signed-off-by: Kees Cook

pstore: Remove filesystem records when backend is unregistered

2020-05-30T17:34:02Z

If a backend was unloaded without having first removed all its associated records in pstorefs, subsequent removals would crash while attempting to call into the now missing backend. Add automatic removal from the tree in pstore_unregister(), so that no references to the backend remain. Reported-by: Luis Henriques Link: https://lore.kernel.org/lkml/87o8yrmv69.fsf@suse.com Link: https://lore.kernel.org/lkml/20200506152114.50375-11-keescook@chromium.org/ Signed-off-by: Kees Cook

pstore: Add locking around superblock changes

2020-05-30T17:33:46Z

Nothing was protecting changes to the pstorefs superblock. Add locking and refactor away is_pstore_mounted(), instead using a helper to add a way to safely lock the pstorefs root inode during filesystem changes. Link: https://lore.kernel.org/lkml/20200506152114.50375-9-keescook@chromium.org/ Signed-off-by: Kees Cook

pstore: Refactor pstorefs record list removal

2020-05-12T16:15:29Z

The "unlink" handling should perform list removal (which can also make sure records don't get double-erased), and the "evict" handling should be responsible only for memory freeing. Link: https://lore.kernel.org/lkml/20200506152114.50375-8-keescook@chromium.org/ Signed-off-by: Kees Cook

pstore: Convert "records_list" locking to mutex

2020-05-12T16:14:18Z

The pstorefs internal list lock doesn't need to be a spinlock and will create problems when trying to access the list in the subsequent patch that will walk the pstorefs records during pstore_unregister(). Change this to a mutex to avoid may_sleep() warnings when unregistering devices. Link: https://lore.kernel.org/lkml/20200506152114.50375-6-keescook@chromium.org/ Signed-off-by: Kees Cook

pstore: Rename "allpstore" to "records_list"

2020-05-12T16:14:05Z

The name "allpstore" doesn't carry much meaning, so rename it to what it actually is: the list of all records present in the filesystem. The lock is also renamed accordingly. Link: https://lore.kernel.org/lkml/20200506152114.50375-5-keescook@chromium.org/ Signed-off-by: Kees Cook

pstore: pstore_ftrace_seq_next should increase position index

2020-02-27T16:04:59Z

In Aug 2018 NeilBrown noticed commit 1f4aace60b0e ("fs/seq_file.c: simplify seq_file iteration code and interface") "Some ->next functions do not increment *pos when they return NULL... Note that such ->next functions are buggy and should be fixed. A simple demonstration is dd if=/proc/swaps bs=1000 skip=1 Choose any block size larger than the size of /proc/swaps. This will always show the whole last line of /proc/swaps" /proc/swaps output was fixed recently, however there are lot of other affected files, and one of them is related to pstore subsystem. If .next function does not change position index, following .show function will repeat output related to current position index. There are at least 2 related problems: - read after lseek beyond end of file, described above by NeilBrown "dd if= bs=1000 skip=1" will generate whole last list - read after lseek on in middle of last line will output expected rest of last line but then repeat whole last line once again. If .show() function generates multy-line output (like pstore_ftrace_seq_show() does ?) following bash script cycles endlessly $ q=;while read -r r;do echo "$((++q)) $r";done < AFFECTED_FILE Unfortunately I'm not familiar enough to pstore subsystem and was unable to find affected pstore-related file on my test node. If .next function does not change position index, following .show function will repeat output related to current position index. Cc: stable@vger.kernel.org Fixes: 1f4aace60b0e ("fs/seq_file.c: simplify seq_file iteration code ...") Link: https://bugzilla.kernel.org/show_bug.cgi?id=206283 Signed-off-by: Vasily Averin Link: https://lore.kernel.org/r/4e49830d-4c88-0171-ee24-1ee540028dad@virtuozzo.com [kees: with robustness tweak from Joel Fernandes ] Signed-off-by: Kees Cook

pstore: Fix double-free in pstore_mkfile() failure path

2019-07-09T04:04:42Z

The pstore_mkfile() function is passed a pointer to a struct pstore_record. On success it consumes this 'record' pointer and references it from the created inode. On failure, however, it may or may not free the record. There are even two different code paths which return -ENOMEM -- one of which does and the other doesn't free the record. Make the behaviour deterministic by never consuming and freeing the record when returning failure, allowing the caller to do the cleanup consistently. Signed-off-by: Norbert Manthey Link: https://lore.kernel.org/r/1562331960-26198-1-git-send-email-nmanthey@amazon.de Fixes: 83f70f0769ddd ("pstore: Do not duplicate record metadata") Fixes: 1dfff7dd67d1a ("pstore: Pass record contents instead of copying") Cc: stable@vger.kernel.org [kees: also move "private" allocation location, rename inode cleanup label] Signed-off-by: Kees Cook