Given corruption in the ftrace records, it might be possible to allocate tmp_prz without assigning prz to it, but still marking it as needing to be freed, which would cause at least a NULL dereference.
smatch warnings: fs/pstore/ram.c:340 ramoops_pstore_read() error: we previously assumed 'prz' could be null (see line 255) https://lists.01.org/pipermail/kbuild-all/2018-December/055528.html Reported-by: Dan Carpenter <dan.carpen...@oracle.com> Fixes: 2fbea82bbb89 ("pstore: Merge per-CPU ftrace records into one") Cc: "Joel Fernandes (Google)" <j...@joelfernandes.org> Signed-off-by: Kees Cook <keesc...@chromium.org> --- fs/pstore/ram.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fs/pstore/ram.c b/fs/pstore/ram.c index e6d9560ea455..96f7d32cd184 100644 --- a/fs/pstore/ram.c +++ b/fs/pstore/ram.c @@ -291,6 +291,7 @@ static ssize_t ramoops_pstore_read(struct pstore_record *record) GFP_KERNEL); if (!tmp_prz) return -ENOMEM; + prz = tmp_prz; free_prz = true; while (cxt->ftrace_read_cnt < cxt->max_ftrace_cnt) { @@ -309,7 +310,6 @@ static ssize_t ramoops_pstore_read(struct pstore_record *record) goto out; } record->id = 0; - prz = tmp_prz; } } -- 2.17.1 -- Kees Cook
