printk: ringbuffer: Fix data block max size check

Currently data_check_size() limits data blocks to a maximum size of
the full buffer minus an ID (long integer):

    max_size <= DATA_SIZE(data_ring) - sizeof(long)

However, this is not an appropriate limit due to the nature of
wrapping data blocks. For example, if a data block is larger than
half the buffer:

    size = (DATA_SIZE(data_ring) / 2) + 8

and begins exactly in the middle of the buffer, then:

    - the data block will wrap
    - the ID will be stored at exactly half of the buffer
    - the record data begins at the beginning of the buffer
    - the record data ends 8 bytes _past_ exactly half of the buffer

The record overwrites itself, i.e. needs more space than the full
buffer!

Luckily printk() is not vulnerable to this problem because
truncate_msg() limits printk-messages to 1/4 of the ringbuffer.
Indeed, by adjusting the printk_ringbuffer KUnit test, which does not
use printk() and its truncate_msg() check, it is easy to see that the
ringbuffer becomes corrupted for records larger than half the buffer
size.

The corruption occurs because data_push_tail() expects it will never
be requested to push the tail beyond the head.

Avoid this problem by adjusting data_check_size() to limit record
sizes to half the buffer size. Also add WARN_ON_ONCE() before
relevant data_push_tail() calls to validate that there are no such
illegal requests. WARN_ON_ONCE() is used, rather than just adding
extra checks to data_push_tail() because it is considered a bug to
attempt such illegal actions.

Link: https://lore.kernel.org/lkml/aMLrGCQSyC8odlFZ@pathway.suse.cz
Signed-off-by: John Ogness <john.ogness@linutronix.de>
Reviewed-by: Petr Mladek <pmladek@suse.com>
Signed-off-by: Petr Mladek <pmladek@suse.com>

0 files changed, 0 insertions, 0 deletions