For devices with multiple input queues, tiqdio_call_inq_handlers()
iterates over all input queues and clears the device's DSCI
during each iteration. If the DSCI is re-armed during one
of the later iterations, we therefore do not scan the previous
queues again.
The re-arming also raises a new adapter interrupt. But its
handler does not trigger a rescan for the device, as the DSCI
has already been erroneously cleared.
This can result in queue stalls on devices with multiple
input queues.
Fix it by clearing the DSCI just once, prior to scanning the queues.
As the code is moved in front of the loop, we also need to access
the DSCI directly (ie irq->dsci) instead of going via each queue's
parent pointer to the same irq. This is not a functional change,
and a follow-up patch will clean up the other users.
In practice, this bug only affects CQ-enabled HiperSockets devices,
ie. devices with sysfs-attribute "hsuid" set. Setting a hsuid is
needed for AF_IUCV socket applications that use HiperSockets
communication.
Fixes: 104ea556ee ("qdio: support asynchronous delivery of storage blocks")
Cc: <stable@vger.kernel.org> # v3.2+
Reviewed-by: Ursula Braun <ubraun@linux.vnet.ibm.com>
Signed-off-by: Julian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
There are three users of adapter interrupts: AP, QDIO and PCI. Each
registers a single adapter interrupt with independent ISCs. Define
a "struct airq" with the interrupt handler, a pointer and a mask for
the local summary indicator and the ISC for the adapter interrupt
source. Convert the indicator array with its fixed number of adapter
interrupt sources per ISE to an array of hlists. This removes the
limitation to 32 adapter interrupts per ISC and allows for arbitrary
memory locations for the local summary indicator.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Move the code to issue the set adapter device controls command to
chsc.c and make it accessible for the qdio code via the wrapper
chsc_sadc.
Reviewed-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Sebastian Ott <sebott@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Now that irq sum accounting for /proc/stat's "intr" line works again we
have the oddity that the sum field (first field) contains only the sum
of the second (external irqs) and third field (I/O interrupts).
The reason for that is that these two fields are already sums of all other
fields. So if we would sum up everything we would count every interrupt
twice.
This is broken since the split interrupt accounting was merged two years
ago: 052ff461c8 "[S390] irq: have detailed
statistics for interrupt types".
To fix this remove the split interrupt fields from /proc/stat's "intr"
line again and only have them in /proc/interrupts.
This restores the old behaviour, seems to be the only sane fix and mimics
a behaviour from other architectures where /proc/interrupts also contains
more than /proc/stat's "intr" line does.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove or replace BUG/BUG_ON where possible and convert WARN_ON
to WARN_ON_ONCE if they can occur freqeuently as pointed out by:
https://lkml.org/lkml/2012/9/27/461
Checks have been removed if:
- the error condition leads to a hardware error which gets logged
and in most cases stops the device
- the error condition is a null pointer access
- the error condition is just pointless or already handled at
another location
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Remove the file name from the comment at top of many files. In most
cases the file name was wrong anyway, so it's rather pointless.
Also unify the IBM copyright statement. We did have a lot of sightly
different statements and wanted to change them one after another
whenever a file gets touched. However that never happened. Instead
people start to take the old/"wrong" statements to use as a template
for new files.
So unify all of them in one go.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
A kernel panic may occur during sending or receiving network packets
on a machine without adapter interrupts since commit d36deae.
The bug is triggered by writing to the shared indicator address which
is set to 0 if the machine doesn't have adapter interrupts.
Make the reading and setting of the shared indicator dependent on the
adapter interrupt feature and while at it move the code to the
file containing the adapter interrupt related code.
Thanks to Jan Jaeger for tracking this down.
Reported-by: Jan Jaeger <jan.jaeger@westnet.com.au>
Tested-by: Jan Jaeger <jan.jaeger@westnet.com.au>
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Add a timestamp per queue and update the timestamp when the queue is
scanned. Add the queue timestamps and the timestamp of the last
adapter interrupt to the debugfs output. The timestamps are useful
for debugging stall conditions.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This patch introduces support for asynchronous delivery of storage blocks for
Hipersockets. Upper layers may exploit this functionality to reuse SBALs for
which the delivery status is still pending.
Signed-off-by: Einar Lueck <elelueck@de.ibm.com>
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This allows us to move duplicated code in <asm/atomic.h>
(atomic_inc_not_zero() for now) to <linux/atomic.h>
Signed-off-by: Arun Sharma <asharma@fb.com>
Reviewed-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: David Miller <davem@davemloft.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The following race can occur with qdio devices that use the shared device
state change indicator:
Device (Shared DSCI) CPU0 CPU1
===============================================================================
1. DSCI 0 => 1,
INT pending
2. Thinint handler
* si_used = 1
* Inbound tasklet_schedule
* DSCI 1 => 0
3. DSCI 0 => 1,
INT pending
4. Thinint handler
* si_used = 1
* Inbound tasklet_schedu
le
=> NOP
5. Inbound tasklet run
6. DSCI = 1,
INT surpressed
7. DSCI 1 => 0
The race would lead to a stall where new data in the input queue is
not recognized so the device stops working in case of no further traffic.
Fix the race by resetting the DSCI before scheduling the inbound tasklet
so the device generates an interrupt if new data arrives in the above
scenario in step 6.
Reviewed-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
If the shared indicator is used the following race leads to
an inbound stall:
Device CPU0 CPU1
========================================================
non-shared DSCI =>1
ALSI => 1
Thin INT
ALSI => 0
non-shared DSCI
tasklets scheduled
shared DSCI => 1
ALSI => 1
shared DSCI => 0
ALSI ? -> set
Thin INT
ALSI => 0
ALSI was set,
shared DSCI => 1
After that no more interrupts occur because the DSCI is still set.
Fix that race by only resetting the shared DSCI if it was actually
set so the tasklets for all shared devices are scheduled and will
run after the interrupt.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Count traditional qdio interrupts and adapter interrupts for qdio
in the interrupt statistics.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The qdio device indicator is freed before the device is notified that
the indicator is reset. This sequence contains a race when the freed
indicator is used by a new device while the reset of the indicator is
still pending. Do the reset operation before freeing the indicator to
avoid that potential race.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Extend the qdio API to allow polling in the upper-layer driver. This
is needed by qeth to use NAPI.
To use the new interface the upper-layer driver must specify the
queue_start_poll(). This callback is used to signal the upper-layer
driver that is has initiative and must process the inbound queue by
calling qdio_get_next_buffers(). If the upper-layer driver wants to
stop polling it calls qdio_start_irq().
Since adapter interrupts are not completely stoppable qdio implements
a software bit QDIO_QUEUE_IRQS_DISABLED to safely disable interrupts for an
input queue.
The old interface is preserved and will be used as is by zfcp.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
The state change indicator is bit 7 not bit 0 of the dsci. Use the
correct bit for setting the indicator.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
percpu.h is included by sched.h and module.h and thus ends up being
included when building most .c files. percpu.h includes slab.h which
in turn includes gfp.h making everything defined by the two files
universally available and complicating inclusion dependencies.
percpu.h -> slab.h dependency is about to be removed. Prepare for
this change by updating users of gfp and slab facilities include those
headers directly instead of assuming availability. As this conversion
needs to touch large number of source files, the following script is
used as the basis of conversion.
http://userweb.kernel.org/~tj/misc/slabh-sweep.py
The script does the followings.
* Scan files for gfp and slab usages and update includes such that
only the necessary includes are there. ie. if only gfp is used,
gfp.h, if slab is used, slab.h.
* When the script inserts a new include, it looks at the include
blocks and try to put the new include such that its order conforms
to its surrounding. It's put in the include block which contains
core kernel includes, in the same order that the rest are ordered -
alphabetical, Christmas tree, rev-Xmas-tree or at the end if there
doesn't seem to be any matching order.
* If the script can't find a place to put a new include (mostly
because the file doesn't have fitting include block), it prints out
an error message indicating which .h file needs to be added to the
file.
The conversion was done in the following steps.
1. The initial automatic conversion of all .c files updated slightly
over 4000 files, deleting around 700 includes and adding ~480 gfp.h
and ~3000 slab.h inclusions. The script emitted errors for ~400
files.
2. Each error was manually checked. Some didn't need the inclusion,
some needed manual addition while adding it to implementation .h or
embedding .c file was more appropriate for others. This step added
inclusions to around 150 files.
3. The script was run again and the output was compared to the edits
from #2 to make sure no file was left behind.
4. Several build tests were done and a couple of problems were fixed.
e.g. lib/decompress_*.c used malloc/free() wrappers around slab
APIs requiring slab.h to be added manually.
5. The script was run on all .h files but without automatically
editing them as sprinkling gfp.h and slab.h inclusions around .h
files could easily lead to inclusion dependency hell. Most gfp.h
inclusion directives were ignored as stuff from gfp.h was usually
wildly available and often used in preprocessor macros. Each
slab.h inclusion directive was examined and added manually as
necessary.
6. percpu.h was updated not to include slab.h.
7. Build test were done on the following configurations and failures
were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my
distributed build env didn't work with gcov compiles) and a few
more options had to be turned off depending on archs to make things
build (like ipr on powerpc/64 which failed due to missing writeq).
* x86 and x86_64 UP and SMP allmodconfig and a custom test config.
* powerpc and powerpc64 SMP allmodconfig
* sparc and sparc64 SMP allmodconfig
* ia64 SMP allmodconfig
* s390 SMP allmodconfig
* alpha SMP allmodconfig
* um on x86_64 SMP allmodconfig
8. percpu.h modifications were reverted so that it could be applied as
a separate patch and serve as bisection point.
Given the fact that I had only a couple of failures from tests on step
6, I'm fairly confident about the coverage of this conversion patch.
If there is a breakage, it's likely to be something in one of the arch
headers which should be easily discoverable easily on most builds of
the specific arch.
Signed-off-by: Tejun Heo <tj@kernel.org>
Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Some parts of cio do not shift PAGE_DEFAULT_KEY correctly and end up
with an incorrect key in their data structures.
Since the default key is zero this doesn't really matter. However if
somebody would use key-controlled protection for debugging purposes
it would be quite helpful if all of this would work as expected.
Also remove a stale declaration.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Revamp the qdio performance statistics and move them from procfs to
debugfs using the seq_file interface. Since the statistics are not
intended for the general user the removal of /proc/qdio_perf should
not surprise anyone.
The per device statistics are disabled by default, writing 1 to
/<debugfs mountpoint>/qdio/<device bus ID>/statistics enables the
statistics for the given device.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Since the adapter interrupt tasklet only schedules the queue tasklets
and contains no code that requires serialization in can be merged
with the adapter interrupt handler. That possibly safes some CPU
cycles.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Move the adapter interrupt tasklet function to the qdio main code
since all the functions used by the tasklet are located there.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The inbound and outbound handlers are nearly identical if the outbound
handler uses first_to_check as end index instead of last_move. Since both
values are identical at that point the handlers can be merged.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
If the qdio module is unloaded the tiqdio tasklet must be terminated
by tasklet_kill. Move the tasklet_kill after the unregistration of
the adapter interrupt so the tiqdio tasklet will not be scheduled
anymore before calling tasklet_kill.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The queue tasklets were stopped with tasklet_disable. Although tasklet_disable
prevents the tasklet from beeing executed it is still possible that a tasklet
is scheduled on a CPU at that point. A following qdio_establish calls
tasklet_init which clears the tasklet count and the tasklet state leading to
the following Oops:
<2>kernel BUG at kernel/softirq.c:392!
<4>illegal operation: 0001 [#1] SMP
<4>Modules linked in: iptable_filter ip_tables x_tables dm_round_robin dm_multipath scsi_dh sg sd_mod crc_t10dif nfs lockd nfs
_acl sunrpc fuse loop dm_mod qeth_l3 ipv6 zfcp qeth scsi_transport_fc qdio scsi_tgt scsi_mod chsc_sch ccwgroup dasd_eckd_mod dasdm
od ext3 mbcache jbd
<4>Supported: Yes
<4>CPU: 0 Not tainted 2.6.27.13-1.1.mz13-default #1
<4>Process blast.LzS_64 (pid: 16445, task: 000000006cc02538, ksp: 000000006cb67998)
<4>Krnl PSW : 0704c00180000000 00000000001399f4 (tasklet_action+0xc8/0x1d4)
<4> R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:3 CC:0 PM:0 EA:3
<4>Krnl GPRS: ffffffff00000030 0000000000000002 0000000000000002 fffffffffffffffe
<4> 000000000013aabe 00000000003b6a18 fffffffffffffffd 0000000000000000
<4> 00000000006705a8 000000007d0914a8 000000007d0914b0 000000007fecfd30
<4> 0000000000000000 00000000003b63e8 000000007fecfd90 000000007fecfd30
<4>Krnl Code: 00000000001399e8: b9200021 cgr %r2,%r1
<4> 00000000001399ec: a7740004 brc 7,1399f4
<4> 00000000001399f0: a7f40001 brc 15,1399f2
<4> >00000000001399f4: c0100027e8ee larl %r1,636bd0
<4> 00000000001399fa: bf1f1008 icm %r1,15,8(%r1)
<4> 00000000001399fe: a7840019 brc 8,139a30
<4> 0000000000139a02: c0300027e8ef larl %r3,636be0
<4> 0000000000139a08: e3c030000004 lg %r12,0(%r3)
<4>Call Trace:
<4>([<0000000000139c12>] tasklet_hi_action+0x112/0x1d4)
<4> [<000000000013aabe>] __do_softirq+0xde/0x1c4
<4> [<000000000010fa2e>] do_softirq+0x96/0xb0
<4> [<000000000013a8d8>] irq_exit+0x70/0xcc
<4> [<000000000010d1d8>] do_extint+0xf0/0x110
<4> [<0000000000113b10>] ext_no_vtime+0x16/0x1a
<4> [<000003e0000a3662>] ext3_dirty_inode+0xe6/0xe8 [ext3]
<4>([<00000000001f6cf2>] __mark_inode_dirty+0x52/0x1d4)
<4> [<000003e0000a44f0>] ext3_ordered_write_end+0x138/0x190 [ext3]
<4> [<000000000018d5ec>] generic_perform_write+0x174/0x230
<4> [<0000000000190144>] generic_file_buffered_write+0xb4/0x194
<4> [<0000000000190864>] __generic_file_aio_write_nolock+0x418/0x454
<4> [<0000000000190ee2>] generic_file_aio_write+0x76/0xe4
<4> [<000003e0000a05c2>] ext3_file_write+0x3e/0xc8 [ext3]
<4> [<00000000001cc2fe>] do_sync_write+0xd6/0x120
<4> [<00000000001ccfc8>] vfs_write+0xac/0x184
<4> [<00000000001cd218>] SyS_write+0x68/0xe0
<4> [<0000000000113402>] sysc_noemu+0x10/0x16
<4> [<0000020000043188>] 0x20000043188
<4>Last Breaking-Event-Address:
<4> [<00000000001399f0>] tasklet_action+0xc4/0x1d4
<6>qdio: 0.0.c61b ZFCP on SC f67 using AI:1 QEBSM:0 PCI:1 TDD:1 SIGA: W AOP
<4> <0>Kernel panic - not syncing: Fatal exception in interrupt
Use tasklet_kill instead of tasklet_disbale. Since tasklet_schedule must not be
called after tasklet_kill use the QDIO_IRQ_STATE_STOPPED to inidicate that a
queue is going down and prevent further tasklet schedules in that case.
Remove superflous tasklet_schedule from input queue setup, at that time
the queues are not ready so the schedule results in a NOP.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Add a mutex to protect the tiq_list. Although reading the list is done
using RCU adding and removing elements from the list must still
happen locked since multiple qdio devices may change the list in parallel
otherwise.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
- Use automatic acknowledgement of incoming buffers in QEBSM mode
- Move ACK for non-QEBSM mode always to the newest buffer to prevent
a race with qdio_stop_polling
- Remove the polling spinlock, the upper layer drivers return new buffers
in the same code path and could not run in parallel
- Don't flood the error log in case of no-target-buffer-empty
- In handle_inbound we check if we would overwrite an ACK'ed buffer, if so
advance the pointer to the oldest ACK'ed buffer so we don't overwrite an
empty buffer in qdio_stop_polling
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
- make qdio_trace a per device view
- remove s390dbf exceptions
- remove CONFIG_QDIO_DEBUG, not needed anymore if we check for the level
before calling sprintf
- use snprintf for dbf entries
- add start markers to see if the dbf view wrapped
- add a global error view for all queues
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
The QEBSM instructions are only available for CONFIG_64BIT, they are not
used under 31 bit. Make compiler happy about the false positive:
drivers/s390/cio/qdio_main.c: In function ?qdio_inbound_q_done?:
drivers/s390/cio/qdio_main.c:532: warning: ?state? may be used uninitialized in this function
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
If qdio_establish fails we call qdio_shutdown to cleanup the
qdio subchannel. The tiq_list entry may not be valid at that
time, therefore we must ignore queues with an invalid list entry
in tiqdio_remove_input_queues.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Julian Wiedmann