elog: Correct behavior when FMAP section doesn't exist on ChromeOS
The elog driver has a really stupid bug that checks a result which is
stored in an unsigned variable for < 0. Surprisingly GCC does not catch
this nonsense right now, and I spent an hour trying out different
warning options without finding one that doesn't also bring a load of
stupid and unavoidable false positives (the biggest offender being
-Wtype-limits, which does exactly what we'd want except for flagging
things like if ((u8)var >= CONFIG_VAR_MIN) where the VAR_MIN Kconfig may
or may not be 0).
So, the only thing we can do is fix this one and wait for the next time
something like that blows up. -.- Also change some more code to make the
behavior more explicit (the old code already intended to work this way
since flash_base is statically initialized to 0, never assigned in the
error path and checked later in elog_init()... but there was an error
message that incorrectly claimed a different fallback behavior, and
explicitly assigning the values makes this easier to see). Finally, add
another state to the elog_initialized variable to avoid trying to
reinitialize a broken eventlog on every event (if it doesn't work the
first time, chances are that it won't work later on during the same boot
either).
BRANCH=None
BUG=chrome-os-partner:35940
TEST=Flashed Jerry with RO 6588.4 and RW 6588.23, observed how it now
cleanly enters recovery mode without blowing its bootblock away with
stray eventlog entries.
Change-Id: I4d93f48d2d01d75a04550d419e023aa42ca95a7a
Signed-off-by: Julius Werner <jwerner@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/243671
Reviewed-by: Duncan Laurie <dlaurie@chromium.org>
1 file changed
