The old code would indefinitely block other users of nf_log_mutex if
a userspace access in proc_dostring() blocked e.g. due to a userfaultfd
region. Fix it by moving proc_dostring() out of the locked region.
This is a followup to commit 266d07cb1c9a ("netfilter: nf_log: fix
sleeping function called from invalid context"), which changed this code
from using rcu_read_lock() to taking nf_log_mutex.
Currently the functions use to check both chip ready and good.
But the chip ready is not enough to check the operation status.
So change this to check the chip good instead of this.
About the retry functions to make sure the error handling remain it.
For the word write functions it is retried for error.
But it is not implemented to retry for the erase functions.
To make sure for the erase functions change to retry as same.
This is needed to prevent the flash erase error caused only once.
It was caused by the error case of chip_good() in the do_erase_oneblock().
Also it was confirmed on the MACRONIX flash device MX29GL512FHT2I-11G.
But the error issue behavior is not able to reproduce at this moment.
The flash controller is parallel Flash interface integrated on BCM53003.
dm_bufio_shrink_count() is called from do_shrink_slab to find out how many
freeable objects are there. The reported value doesn't have to be precise,
so we don't need to take the dm-bufio lock.
The v21 version of the NAND flash controller contains a Spare Area Size
Register (SPAS) at offset 0x10. Its setting defaults to the maximum
spare area size of 218 bytes. The size that is set in this register is
used by the controller when it calculates the ECC bytes internally in
hardware.
Usually, this register is updated from settings in the IIM fuses when
the system is booting from NAND flash. For other boot media, however,
the SPAS register remains at the default setting, which may not work for
the particular flash chip on the board. The same goes for flash chips
whose configuration cannot be set in the IIM fuses (e.g. chips with 2k
sector size and 128 bytes spare area size can't be configured in the IIM
fuses on imx25 systems).
Set the SPAS register explicitly during the preset operation. Derive the
register value from mtd->oobsize that was detected during probe by
decoding the flash chip's ID bytes.
While at it, rename the define for the spare area register's offset to
NFC_V21_RSLTSPARE_AREA. The register at offset 0x10 on v1 controllers is
different from the register on v21 controllers.
Fixes: d484018 ("mtd: mxc_nand: set NFC registers after reset")
Cc: <email address hidden>
Signed-off-by: Martin Kaiser <email address hidden>
Reviewed-by: Sascha Hauer <email address hidden>
Reviewed-by: Miquel Raynal <email address hidden>
Signed-off-by: Boris Brezillon <email address hidden>
Signed-off-by: Greg Kroah-Hartman <email address hidden>
Signed-off-by: Kleber Sacilotto de Souza <email address hidden>
ad5854a...
by
Mikulas Patocka <email address hidden>
dm bufio: drop the lock when doing GFP_NOIO allocation
If the first allocation attempt using GFP_NOWAIT fails, drop the lock
and retry using GFP_NOIO allocation (lock is dropped because the
allocation can take some time).
Note that we won't do GFP_NOIO allocation when we loop for the second
time, because the lock shouldn't be dropped between __wait_for_free_buffer
and __get_unclaimed_buffer.
...and that looks to be the one holding the mutex.
The problem has been reproduced on fairly easily:
0. Be running Chrome OS w/ verity enabled on the root filesystem
1. Pick test patch: http://crosreview.com/412360
2. Install launchBalloons.sh and balloon.arm from http://crbug.com/468342
...that's just a memory stress test app.
3. On a 4GB rk3399 machine, run
nice ./launchBalloons.sh 4 900 100000
...that tries to eat 4 * 900 MB of memory and keep accessing.
4. Login to the Chrome web browser and restore many tabs
With that, I've seen printouts like:
DOUG: long bufio 90758 ms
...and stack trace always show's we're in dm_bufio_prefetch().
The problem is that we try to allocate memory with GFP_NOIO while
we're holding the dm_bufio lock. Instead we should be using
GFP_NOWAIT. Using GFP_NOIO can cause us to sleep while holding the
lock and that causes the above problems.
The current behavior explained by David Rientjes:
It will still try reclaim initially because __GFP_WAIT (or
__GFP_KSWAPD_RECLAIM) is set by GFP_NOIO. This is the cause of
contention on dm_bufio_lock() that the thread holds. You want to
pass GFP_NOWAIT instead of GFP_NOIO to alloc_buffer() when holding a
mutex that can be contended by a concurrent slab shrinker (if
count_objects didn't use a trylock, this pattern would trivially
deadlock).
This change significantly increases responsiveness of the system while
in this state. It makes a real difference because it unblocks kswapd.
In the bug report analyzed, kswapd was hung: