| Age | Commit message (Collapse) | Author |
|---|
|
Finn Thain says:
====================
Fixes for SONIC ethernet driver
Various SONIC driver problems have become apparent over the years,
including tx watchdog timeouts, lost packets and duplicated packets.
The problems are mostly caused by bugs in buffer handling, locking and
(re-)initialization code.
This patch series resolves these problems.
This series has been tested on National Semiconductor hardware (macsonic),
qemu-system-m68k (macsonic) and qemu-system-mips64el (jazzsonic).
The emulated dp8393x device used in QEMU also has bugs.
I have fixed the bugs that I know of in a series of patches at,
https://github.com/fthain/qemu/commits/sonic
Changed since v1:
- Minor revisions as described in commit logs.
- Deferred net-next patches.
Changed since v2:
- Minor revisions as described in commit logs.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Section 5.5.3.2 of the datasheet says,
If FIFO Underrun, Byte Count Mismatch, Excessive Collision, or
Excessive Deferral (if enabled) errors occur, transmission ceases.
In this situation, the chip asserts a TXER interrupt rather than TXDN.
But the handler for the TXDN is the only way that the transmit queue
gets restarted. Hence, an aborted transmission can result in a watchdog
timeout.
This problem can be reproduced on congested link, as that can result in
excessive transmitter collisions. Another way to reproduce this is with
a FIFO Underrun, which may be caused by DMA latency.
In event of a TXER interrupt, prevent a watchdog timeout by restarting
transmission.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Section 4.3.1 of the datasheet says,
This bit [TXP] must not be set if a Load CAM operation is in
progress (LCAM is set). The SONIC will lock up if both bits are
set simultaneously.
Testing has shown that the driver sometimes attempts to set LCAM
while TXP is set. Avoid this by waiting for command completion
before and after giving the LCAM command.
After issuing the Load CAM command, poll for !SONIC_CR_LCAM rather than
SONIC_INT_LCD, because the SONIC_CR_TXP bit can't be used until
!SONIC_CR_LCAM.
When in reset mode, take the opportunity to reset the CAM Enable
register.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
There are several issues relating to command register usage during
chip initialization.
Firstly, the SONIC sometimes comes out of software reset with the
Start Timer bit set. This gets logged as,
macsonic macsonic eth0: sonic_init: status=24, i=101
Avoid this by giving the Stop Timer command earlier than later.
Secondly, the loop that waits for the Read RRA command to complete has
the break condition inverted. That's why the for loop iterates until
its termination condition. Call the helper for this instead.
Finally, give the Receiver Enable command after clearing interrupts,
not before, to avoid the possibility of losing an interrupt.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Make sure the SONIC's DMA engine is idle before altering the transmit
and receive descriptors. Add a helper for this as it will be needed
again.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
As soon as the driver is finished with a receive buffer it allocs a new
one and overwrites the corresponding RRA entry with a new buffer pointer.
Problem is, the buffer pointer is split across two word-sized registers.
It can't be updated in one atomic store. So this operation races with the
chip while it stores received packets and advances its RRP register.
This could result in memory corruption by a DMA write.
Avoid this problem by adding buffers only at the location given by the
RWP register, in accordance with the National Semiconductor datasheet.
Re-factor this code into separate functions to calculate a RRA pointer
and to update the RWP.
Fixes: efcce839360f ("[PATCH] macsonic/jazzsonic network drivers update")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
After sonic_tx_timeout() calls sonic_init(), it can happen that
sonic_rx() will subsequently encounter a receive descriptor with no
flags set. Remove the comment that says that this can't happen.
When giving a receive descriptor to the SONIC, clear the descriptor
status field. That way, any rx descriptor with flags set can only be
a newly received packet.
Don't process a descriptor without the LPKT bit set. The buffer is
still in use by the SONIC.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The while loop in sonic_rx() traverses the rx descriptor ring. It stops
when it reaches a descriptor that the SONIC has not used. Each iteration
advances the EOL flag so the SONIC can keep using more descriptors.
Therefore, the while loop has no definite termination condition.
The algorithm described in the National Semiconductor literature is quite
different. It consumes descriptors up to the one with its EOL flag set
(which will also have its "in use" flag set). All freed descriptors are
then returned to the ring at once, by adjusting the EOL flags (and link
pointers).
Adopt the algorithm from datasheet as it's simpler, terminates quickly
and avoids a lot of pointless descriptor EOL flag changes.
Fixes: efcce839360f ("[PATCH] macsonic/jazzsonic network drivers update")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The SONIC can sometimes advance its rx buffer pointer (RRP register)
without advancing its rx descriptor pointer (CRDA register). As a result
the index of the current rx descriptor may not equal that of the current
rx buffer. The driver mistakenly assumes that they are always equal.
This assumption leads to incorrect packet lengths and possible packet
duplication. Avoid this by calling a new function to locate the buffer
corresponding to a given descriptor.
Fixes: efcce839360f ("[PATCH] macsonic/jazzsonic network drivers update")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The tx_aborted_errors statistic should count packets flagged with EXD,
EXC, FU, or BCM bits because those bits denote an aborted transmission.
That corresponds to the bitmask 0x0446, not 0x0642. Use macros for these
constants to avoid mistakes. Better to leave out FIFO Underruns (FU) as
there's a separate counter for that purpose.
Don't lump all these errors in with the general tx_errors counter as
that's used for tx timeout events.
On the rx side, don't count RDE and RBAE interrupts as dropped packets.
These interrupts don't indicate a lost packet, just a lack of resources.
When a lack of resources results in a lost packet, this gets reported
in the rx_missed_errors counter (along with RFO events).
Don't double-count rx_frame_errors and rx_crc_errors.
Don't use the general rx_errors counter for events that already have
special counters.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The driver accesses descriptor memory which is simultaneously accessed by
the chip, so the compiler must not be allowed to re-order CPU accesses.
sonic_buf_get() used 'volatile' to prevent that. sonic_buf_put() should
have done so too but was overlooked.
Fixes: efcce839360f ("[PATCH] macsonic/jazzsonic network drivers update")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The chip can change a packet's descriptor status flags at any time.
However, an active interrupt flag gets cleared rather late. This
allows a race condition that could theoretically lose an interrupt.
Fix this by clearing asserted interrupt flags immediately.
Fixes: efcce839360f ("[PATCH] macsonic/jazzsonic network drivers update")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
The netif_stop_queue() call in sonic_send_packet() races with the
netif_wake_queue() call in sonic_interrupt(). This causes issues
like "NETDEV WATCHDOG: eth0 (macsonic): transmit queue 0 timed out".
Fix this by disabling interrupts when accessing tx_skb[] and next_tx.
Update a comment to clarify the synchronization properties.
Fixes: efcce839360f ("[PATCH] macsonic/jazzsonic network drivers update")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@telegraphics.com.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
As the only 10G PHY interface type defined at the moment the code
was developed was XGMII, although the PHY interface mode used was
not XGMII, XGMII was used in the code to denote 10G. This patch
renames the 10G interface mode to remove the ambiguity.
Signed-off-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Madalin Bucur says:
====================
net: fsl/fman: address erratum A011043
This addresses a HW erratum on some QorIQ DPAA devices.
MDIO reads to internal PCS registers may result in having
the MDIO_CFG[MDIO_RD_ER] bit set, even when there is no
error and read data (MDIO_DATA[MDIO_DATA]) is correct.
Software may get false read error when reading internal
PCS registers through MDIO. As a workaround, all internal
MDIO accesses should ignore the MDIO_CFG[MDIO_RD_ER] bit.
When the issue was present, one could see such errors
during boot:
mdio_bus ffe4e5000: Error while reading PHY0 reg at 3.3
====================
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
When fsl,erratum-a011043 is set, adjust for erratum A011043:
MDIO reads to internal PCS registers may result in having
the MDIO_CFG[MDIO_RD_ER] bit set, even when there is no
error and read data (MDIO_DATA[MDIO_DATA]) is correct.
Software may get false read error when reading internal
PCS registers through MDIO. As a workaround, all internal
MDIO accesses should ignore the MDIO_CFG[MDIO_RD_ER] bit.
Signed-off-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add fsl,erratum-a011043 to internal MDIO buses.
Software may get false read error when reading internal
PCS registers through MDIO. As a workaround, all internal
MDIO accesses should ignore the MDIO_CFG[MDIO_RD_ER] bit.
Signed-off-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Add an entry for erratum A011043: the MDIO_CFG[MDIO_RD_ER]
bit may be falsely set when reading internal PCS registers.
MDIO reads to internal PCS registers may result in having
the MDIO_CFG[MDIO_RD_ER] bit set, even when there is no
error and read data (MDIO_DATA[MDIO_DATA]) is correct.
Software may get false read error when reading internal
PCS registers through MDIO. As a workaround, all internal
MDIO accesses should ignore the MDIO_CFG[MDIO_RD_ER] bit.
Signed-off-by: Madalin Bucur <madalin.bucur@oss.nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Driver while collecting firmware dump takes longer time to
collect/process some of the firmware dump entries/memories.
Bigger capture masks makes it worse as it results in larger
amount of data being collected and results in CPU soft lockup.
Place cond_resched() in some of the driver flows that are
expectedly time consuming to relinquish the CPU to avoid CPU
soft lockup panic.
Signed-off-by: Shahed Shaikh <shshaikh@marvell.com>
Tested-by: Yonggen Xu <Yonggen.Xu@dell.com>
Signed-off-by: Manish Chopra <manishc@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
|
|
Pull XArray fixes from Matthew Wilcox:
"Primarily bugfixes, mostly around handling index wrap-around
correctly.
A couple of doc fixes and adding missing APIs.
I had an oops live on stage at linux.conf.au this year, and it turned
out to be a bug in xas_find() which I can't prove isn't triggerable in
the current codebase. Then in looking for the bug, I spotted two more
bugs.
The bots have had a few days to chew on this with no problems
reported, and it passes the test-suite (which now has more tests to
make sure these problems don't come back)"
* tag 'xarray-5.5' of git://git.infradead.org/users/willy/linux-dax:
XArray: Add xa_for_each_range
XArray: Fix xas_find returning too many entries
XArray: Fix xa_find_after with multi-index entries
XArray: Fix infinite loop with entry at ULONG_MAX
XArray: Add wrappers for nested spinlocks
XArray: Improve documentation of search marks
XArray: Fix xas_pause at ULONG_MAX
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing fixes from Steven Rostedt:
"Various tracing fixes:
- Fix a function comparison warning for a xen trace event macro
- Fix a double perf_event linking to a trace_uprobe_filter for
multiple events
- Fix suspicious RCU warnings in trace event code for using
list_for_each_entry_rcu() when the "_rcu" portion wasn't needed.
- Fix a bug in the histogram code when using the same variable
- Fix a NULL pointer dereference when tracefs lockdown enabled and
calling trace_set_default_clock()
- A fix to a bug found with the double perf_event linking patch"
* tag 'trace-v5.5-rc6-2' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
tracing/uprobe: Fix to make trace_uprobe_filter alignment safe
tracing: Do not set trace clock if tracefs lockdown is in effect
tracing: Fix histogram code when expression has same var as value
tracing: trigger: Replace unneeded RCU-list traversals
tracing/uprobe: Fix double perf_event linking on multiprobe uprobe
tracing: xen: Ordered comparison of function pointers
|
|
Pull ceph fix from Ilya Dryomov:
"A fix for a potential use-after-free from Jeff, marked for stable"
* tag 'ceph-for-5.5-rc8' of https://github.com/ceph/ceph-client:
ceph: hold extra reference to r_parent over life of request
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Pull power management fix from Rafael Wysocki:
"Prevent the kernel from crashing during resume from hibernation if
free pages contain leftover data from the restore kernel and
init_on_free is set (Alexander Potapenko)"
* tag 'pm-5.5-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
PM: hibernate: fix crashes with init_on_free=1
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Pull PCI fix from Bjorn Helgaas:
"Mark ATS as broken on AMD Navi14 GPU rev 0xc5 (Alex Deucher)"
* tag 'pci-v5.5-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
PCI: Mark AMD Navi14 GPU rev 0xc5 ATS as broken
|
|
There is a spelling mistake in a ldm_error message. Fix it.
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
When shrink btree node cache from c->btree_cache in bch_mca_scan(),
no matter the selected node is reaped or not, it will be rotated from
the head to the tail of c->btree_cache list. But in bcache journal
code, when flushing the btree nodes with oldest journal entry, btree
nodes are iterated and slected from the tail of c->btree_cache list in
btree_flush_write(). The list_rotate_left() in bch_mca_scan() will
make btree_flush_write() iterate more nodes in c->btree_list in reverse
order.
This patch just reaps the selected btree node cache, and not move it
from the head to the tail of c->btree_cache list. Then bch_mca_scan()
will not mess up c->btree_cache list to btree_flush_write().
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
In order to skip the most recently freed btree node cahce, currently
in bch_mca_scan() the first 3 caches in c->btree_cache_freeable list
are skipped when shrinking bcache node caches in bch_mca_scan(). The
related code in bch_mca_scan() is,
737 list_for_each_entry_safe(b, t, &c->btree_cache_freeable, list) {
738 if (nr <= 0)
739 goto out;
740
741 if (++i > 3 &&
742 !mca_reap(b, 0, false)) {
lines free cache memory
746 }
747 nr--;
748 }
The problem is, if virtual memory code calls bch_mca_scan() and
the calculated 'nr' is 1 or 2, then in the above loop, nothing will
be shunk. In such case, if slub/slab manager calls bch_mca_scan()
for many times with small scan number, it does not help to shrink
cache memory and just wasts CPU cycles.
This patch just selects btree node caches from tail of the
c->btree_cache_freeable list, then the newly freed host cache can
still be allocated by mca_alloc(), and at least 1 node can be shunk.
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The member 'accessed' of struct btree is used in bch_mca_scan() when
shrinking btree node caches. The original idea is, if b->accessed is
set, clean it and look at next btree node cache from c->btree_cache
list, and only shrink the caches whose b->accessed is cleaned. Then
only cold btree node cache will be shrunk.
But when I/O pressure is high, it is very probably that b->accessed
of a btree node cache will be set again in bch_btree_node_get()
before bch_mca_scan() selects it again. Then there is no chance for
bch_mca_scan() to shrink enough memory back to slub or slab system.
This patch removes member accessed from struct btree, then once a
btree node ache is selected, it will be immediately shunk. By this
change, bch_mca_scan() may release btree node cahce more efficiently.
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
It's useful to dump written block and keys on btree write, this patch
add them into trace_bcache_btree_write.
Signed-off-by: Guoju Fang <fangguoju@gmail.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
the commit 91be66e1318f ("bcache: performance improvement for
btree_flush_write()") was an effort to flushing btree node with oldest
btree node faster in following methods,
- Only iterate dirty btree nodes in c->btree_cache, avoid scanning a lot
of clean btree nodes.
- Take c->btree_cache as a LRU-like list, aggressively flushing all
dirty nodes from tail of c->btree_cache util the btree node with
oldest journal entry is flushed. This is to reduce the time of holding
c->bucket_lock.
Guoju Fang and Shuang Li reported that they observe unexptected extra
write I/Os on cache device after applying the above patch. Guoju Fang
provideed more detailed diagnose information that the aggressive
btree nodes flushing may cause 10x more btree nodes to flush in his
workload. He points out when system memory is large enough to hold all
btree nodes in memory, c->btree_cache is not a LRU-like list any more.
Then the btree node with oldest journal entry is very probably not-
close to the tail of c->btree_cache list. In such situation much more
dirty btree nodes will be aggressively flushed before the target node
is flushed. When slow SATA SSD is used as cache device, such over-
aggressive flushing behavior will cause performance regression.
After spending a lot of time on debug and diagnose, I find the real
condition is more complicated, aggressive flushing dirty btree nodes
from tail of c->btree_cache list is not a good solution.
- When all btree nodes are cached in memory, c->btree_cache is not
a LRU-like list, the btree nodes with oldest journal entry won't
be close to the tail of the list.
- There can be hundreds dirty btree nodes reference the oldest journal
entry, before flushing all the nodes the oldest journal entry cannot
be reclaimed.
When the above two conditions mixed together, a simply flushing from
tail of c->btree_cache list is really NOT a good idea.
Fortunately there is still chance to make btree_flush_write() work
better. Here is how this patch avoids unnecessary btree nodes flushing,
- Only acquire c->journal.lock when getting oldest journal entry of
fifo c->journal.pin. In rested locations check the journal entries
locklessly, so their values can be changed on other cores
in parallel.
- In loop list_for_each_entry_safe_reverse(), checking latest front
point of fifo c->journal.pin. If it is different from the original
point which we get with locking c->journal.lock, it means the oldest
journal entry is reclaim on other cores. At this moment, all selected
dirty nodes recorded in array btree_nodes[] are all flushed and clean
on other CPU cores, it is unncessary to iterate c->btree_cache any
longer. Just quit the list_for_each_entry_safe_reverse() loop and
the following for-loop will skip all the selected clean nodes.
- Find a proper time to quit the list_for_each_entry_safe_reverse()
loop. Check the refcount value of orignial fifo front point, if the
value is larger than selected node number of btree_nodes[], it means
more matching btree nodes should be scanned. Otherwise it means no
more matching btee nodes in rest of c->btree_cache list, the loop
can be quit. If the original oldest journal entry is reclaimed and
fifo front point is updated, the refcount of original fifo front point
will be 0, then the loop will be quit too.
- Not hold c->bucket_lock too long time. c->bucket_lock is also required
for space allocation for cached data, hold it for too long time will
block regular I/O requests. When iterating list c->btree_cache, even
there are a lot of maching btree nodes, in order to not holding
c->bucket_lock for too long time, only BTREE_FLUSH_NR nodes are
selected and to flush in following for-loop.
With this patch, only btree nodes referencing oldest journal entry
are flushed to cache device, no aggressive flushing for unnecessary
btree node any more. And in order to avoid blocking regluar I/O
requests, each time when btree_flush_write() called, at most only
BTREE_FLUSH_NR btree nodes are selected to flush, even there are more
maching btree nodes in list c->btree_cache.
At last, one more thing to explain: Why it is safe to read front point
of c->journal.pin without holding c->journal.lock inside the
list_for_each_entry_safe_reverse() loop ?
Here is my answer: When reading the front point of fifo c->journal.pin,
we don't need to know the exact value of front point, we just want to
check whether the value is different from the original front point
(which is accurate value because we get it while c->jouranl.lock is
held). For such purpose, it works as expected without holding
c->journal.lock. Even the front point is changed on other CPU core and
not updated to local core, and current iterating btree node has
identical journal entry local as original fetched fifo front point, it
is still safe. Because after holding mutex b->write_lock (with memory
barrier) this btree node can be found as clean and skipped, the loop
will quite latter when iterate on next node of list c->btree_cache.
Fixes: 91be66e1318f ("bcache: performance improvement for btree_flush_write()")
Reported-by: Guoju Fang <fangguoju@gmail.com>
Reported-by: Shuang Li <psymon@bonuscloud.io>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
To explain the pages allocated from mempool state->pool can be
swapped in __btree_sort(), because state->pool is a page pool,
which allocates pages by alloc_pages() indeed.
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The crc64_be() is declared in <linux/crc64.h> so include
this where the symbol is defined to avoid the following
warning:
lib/crc64.c:43:12: warning: symbol 'crc64_be' was not declared. Should it be static?
Signed-off-by: Ben Dooks (Codethink) <ben.dooks@codethink.co.uk>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Avoid a pointless dependency on buffer heads in bcache by simply open
coding reading a single page. Also add a SB_OFFSET define for the
byte offset of the superblock instead of using magic numbers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
This allows to properly build the superblock bio including the offset in
the page using the normal bio helpers. This fixes writing the superblock
for page sizes larger than 4k where the sb write bio would need an offset
in the bio_vec.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Returning the properly typed actual data structure insteaf of the
containing struct page will save the callers some work going
forward.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Avoid an extra reference count roundtrip by transferring the sb_page
ownership to the lower level register helpers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The patch "bcache: rework error unwinding in register_bcache" introduces
a use-after-free regression in register_bcache(). Here are current code,
2510 out_free_path:
2511 kfree(path);
2512 out_module_put:
2513 module_put(THIS_MODULE);
2514 out:
2515 pr_info("error %s: %s", path, err);
2516 return ret;
If some error happens and the above code path is executed, at line 2511
path is released, but referenced at line 2515. Then KASAN reports a use-
after-free error message.
This patch changes line 2515 in the following way to fix the problem,
2515 pr_info("error %s: %s", path?path:"", err);
Signed-off-by: Coly Li <colyli@suse.de>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Patch "bcache: rework error unwinding in register_bcache" from
Christoph Hellwig changes the local variables 'path' and 'err'
in undefined initial state. If the code in register_bcache() jumps
to label 'out:' or 'out_module_put:' by goto, these two variables
might be reference with undefined value by the following line,
out_module_put:
module_put(THIS_MODULE);
out:
pr_info("error %s: %s", path, err);
return ret;
Therefore this patch initializes these two local variables properly
in register_bcache() to avoid such issue.
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Split the successful and error return path, and use one goto label for each
resource to unwind. This also fixes some small errors like leaking the
module reference count in the reboot case (which seems entirely harmless)
or printing the wrong warning messages for early failures.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Split out an on-disk version struct cache_sb with the proper endianness
annotations. This fixes a fair chunk of sparse warnings, but there are
some left due to the way the checksum is defined.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Same as cache device, the buffer page needs to be put while
freeing cached_dev. Otherwise a page would be leaked every
time a cached_dev is stopped.
Signed-off-by: Liang Chen <liangchen.linux@gmail.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Coly Li <colyli@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
In commit 9f79b78ef744 ("Convert filldir[64]() from __put_user() to
unsafe_put_user()") I changed filldir to not do individual __put_user()
accesses, but instead use unsafe_put_user() surrounded by the proper
user_access_begin/end() pair.
That make them enormously faster on modern x86, where the STAC/CLAC
games make individual user accesses fairly heavy-weight.
However, the user_access_begin() range was not really the exact right
one, since filldir() has the unfortunate problem that it needs to not
only fill out the new directory entry, it also needs to fix up the
previous one to contain the proper file offset.
It's unfortunate, but the "d_off" field in "struct dirent" is _not_ the
file offset of the directory entry itself - it's the offset of the next
one. So we end up backfilling the offset in the previous entry as we
walk along.
But since x86 didn't really care about the exact range, and used to be
the only architecture that did anything fancy in user_access_begin() to
begin with, the filldir[64]() changes did something lazy, and even
commented on it:
/*
* Note! This range-checks 'previous' (which may be NULL).
* The real range was checked in getdents
*/
if (!user_access_begin(dirent, sizeof(*dirent)))
goto efault;
and it all worked fine.
But now 32-bit ppc is starting to also implement user_access_begin(),
and the fact that we faked the range to only be the (possibly not even
valid) previous directory entry becomes a problem, because ppc32 will
actually be using the range that is passed in for more than just "check
that it's user space".
This is a complete rewrite of Christophe's original patch.
By saving off the record length of the previous entry instead of a
pointer to it in the filldir data structures, we can simplify the range
check and the writing of the previous entry d_off field. No need for
any conditionals in the user accesses themselves, although we retain the
conditional EINTR checking for the "was this the first directory entry"
signal handling latency logic.
Fixes: 9f79b78ef744 ("Convert filldir[64]() from __put_user() to unsafe_put_user()")
Link: https://lore.kernel.org/lkml/a02d3426f93f7eb04960a4d9140902d278cab0bb.1579697910.git.christophe.leroy@c-s.fr/
Link: https://lore.kernel.org/lkml/408c90c4068b00ea8f1c41cca45b84ec23d4946b.1579783936.git.christophe.leroy@c-s.fr/
Reported-and-tested-by: Christophe Leroy <christophe.leroy@c-s.fr>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Commit 8a23eb804ca4 ("Make filldir[64]() verify the directory entry
filename is valid") added some minimal validity checks on the directory
entries passed to filldir[64](). But they really were pretty minimal.
This fleshes out at least the name length check: we used to disallow
zero-length names, but really, negative lengths or oevr-long names
aren't ok either. Both could happen if there is some filesystem
corruption going on.
Now, most filesystems tend to use just an "unsigned char" or similar for
the length of a directory entry name, so even with a corrupt filesystem
you should never see anything odd like that. But since we then use the
name length to create the directory entry record length, let's make sure
it actually is half-way sensible.
Note how POSIX states that the size of a path component is limited by
NAME_MAX, but we actually use PATH_MAX for the check here. That's
because while NAME_MAX is generally the correct maximum name length
(it's 255, for the same old "name length is usually just a byte on
disk"), there's nothing in the VFS layer that really cares.
So the real limitation at a VFS layer is the total pathname length you
can pass as a filename: PATH_MAX.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Since btrfs was migrated to use the generic VFS helpers for clone and
deduplication, it stopped allowing for the last block of a file to be
deduplicated when the source file size is not sector size aligned (when
eof is somewhere in the middle of the last block). There are two reasons
for that:
1) The generic code always rounds down, to a multiple of the block size,
the range's length for deduplications. This means we end up never
deduplicating the last block when the eof is not block size aligned,
even for the safe case where the destination range's end offset matches
the destination file's size. That rounding down operation is done at
generic_remap_check_len();
2) Because of that, the btrfs specific code does not expect anymore any
non-aligned range length's for deduplication and therefore does not
work if such nona-aligned length is given.
This patch addresses that second part, and it depends on a patch that
fixes generic_remap_check_len(), in the VFS, which was submitted ealier
and has the following subject:
"fs: allow deduplication of eof block into the end of the destination file"
These two patches address reports from users that started seeing lower
deduplication rates due to the last block never being deduplicated when
the file size is not aligned to the filesystem's block size.
Link: https://lore.kernel.org/linux-btrfs/2019-1576167349.500456@svIo.N5dq.dFFD/
CC: stable@vger.kernel.org # 5.1+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
We always round down, to a multiple of the filesystem's block size, the
length to deduplicate at generic_remap_check_len(). However this is only
needed if an attempt to deduplicate the last block into the middle of the
destination file is requested, since that leads into a corruption if the
length of the source file is not block size aligned. When an attempt to
deduplicate the last block into the end of the destination file is
requested, we should allow it because it is safe to do it - there's no
stale data exposure and we are prepared to compare the data ranges for
a length not aligned to the block (or page) size - in fact we even do
the data compare before adjusting the deduplication length.
After btrfs was updated to use the generic helpers from VFS (by commit
34a28e3d77535e ("Btrfs: use generic_remap_file_range_prep() for cloning
and deduplication")) we started to have user reports of deduplication
not reflinking the last block anymore, and whence users getting lower
deduplication scores. The main use case is deduplication of entire
files that have a size not aligned to the block size of the filesystem.
We already allow cloning the last block to the end (and beyond) of the
destination file, so allow for deduplication as well.
Link: https://lore.kernel.org/linux-btrfs/2019-1576167349.500456@svIo.N5dq.dFFD/
CC: stable@vger.kernel.org # 5.1+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Should work properly with the latest sbios on 5.5 and newer
kernels.
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Sometimes when running generic/475 we would trip the
WARN_ON(cache->reserved) check when free'ing the block groups on umount.
This is because sometimes we don't commit the transaction because of IO
errors and thus do not cleanup the tree logs until at umount time.
These blocks are still reserved until they are cleaned up, but they
aren't cleaned up until _after_ we do the free block groups work. Fix
this by moving the free after free'ing the fs roots, that way all of the
tree logs are cleaned up and we have a properly cleaned fs. A bunch of
loops of generic/475 confirmed this fixes the problem.
CC: stable@vger.kernel.org # 4.9+
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Current code doesn't correctly handle the situation which arises when
a file system that has METADATA_UUID_INCOMPAT flag set and has its FSID
changed to the one in metadata uuid. This causes the incompat flag to
disappear.
In case of a power failure we could end up in a situation where part of
the disks in a multi-disk filesystem are correctly reverted to
METADATA_UUID_INCOMPAT flag unset state, while others have
METADATA_UUID_INCOMPAT set and CHANGING_FSID_V2_IN_PROGRESS.
This patch corrects the behavior required to handle the case where a
disk of the second type is scanned first, creating the necessary
btrfs_fs_devices. Subsequently, when a disk which has already completed
the transition is scanned it should overwrite the data in
btrfs_fs_devices.
Reported-by: Su Yue <Damenly_Su@gmx.com>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
There is one more cases which isn't handled by the original metadata
uuid work. Namely, when a filesystem has METADATA_UUID incompat bit and
the user decides to change the FSID to the original one e.g. have
metadata_uuid and fsid match. In case of power failure while this
operation is in progress we could end up in a situation where some of
the disks have the incompat bit removed and the other half have both
METADATA_UUID_INCOMPAT and FSID_CHANGING_IN_PROGRESS flags.
This patch handles the case where a disk that has successfully changed
its FSID such that it equals METADATA_UUID is scanned first.
Subsequently when a disk with both
METADATA_UUID_INCOMPAT/FSID_CHANGING_IN_PROGRESS flags is scanned
find_fsid_changed won't be able to find an appropriate btrfs_fs_devices.
This is done by extending find_fsid_changed to correctly find
btrfs_fs_devices whose metadata_uuid/fsid are the same and they match
the metadata_uuid of the currently scanned device.
Fixes: cc5de4e70256 ("btrfs: Handle final split-brain possibility during fsid change")
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reported-by: Su Yue <Damenly_Su@gmx.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
find_fsid became rather hairy with the introduction of metadata uuid
changing feature. Alleviate this by factoring out the metadata uuid
specific code in a dedicated function which deals with finding
correct fsid for a device with changed uuid.
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Su Yue <Damenly_Su@gmx.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
