Maintain a per-CPU global "struct pt_regs *" variable which can be used instead
of passing regs around manually through all ~1800 interrupt handlers in the
Linux kernel.
The regs pointer is used in few places, but it potentially costs both stack
space and code to pass it around. On the FRV arch, removing the regs parameter
from all the genirq function results in a 20% speed up of the IRQ exit path
(ie: from leaving timer_interrupt() to leaving do_IRQ()).
Where appropriate, an arch may override the generic storage facility and do
something different with the variable. On FRV, for instance, the address is
maintained in GR28 at all times inside the kernel as part of general exception
handling.
Having looked over the code, it appears that the parameter may be handed down
through up to twenty or so layers of functions. Consider a USB character
device attached to a USB hub, attached to a USB controller that posts its
interrupts through a cascaded auxiliary interrupt controller. A character
device driver may want to pass regs to the sysrq handler through the input
layer which adds another few layers of parameter passing.
I've build this code with allyesconfig for x86_64 and i386. I've runtested the
main part of the code on FRV and i386, though I can't test most of the drivers.
I've also done partial conversion for powerpc and MIPS - these at least compile
with minimal configurations.
This will affect all archs. Mostly the changes should be relatively easy.
Take do_IRQ(), store the regs pointer at the beginning, saving the old one:
struct pt_regs *old_regs = set_irq_regs(regs);
And put the old one back at the end:
set_irq_regs(old_regs);
Don't pass regs through to generic_handle_irq() or __do_IRQ().
In timer_interrupt(), this sort of change will be necessary:
- update_process_times(user_mode(regs));
- profile_tick(CPU_PROFILING, regs);
+ update_process_times(user_mode(get_irq_regs()));
+ profile_tick(CPU_PROFILING);
I'd like to move update_process_times()'s use of get_irq_regs() into itself,
except that i386, alone of the archs, uses something other than user_mode().
Some notes on the interrupt handling in the drivers:
(*) input_dev() is now gone entirely. The regs pointer is no longer stored in
the input_dev struct.
(*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does
something different depending on whether it's been supplied with a regs
pointer or not.
(*) Various IRQ handler function pointers have been moved to type
irq_handler_t.
Signed-Off-By: David Howells <dhowells@redhat.com>
(cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6: (48 commits)
ieee1394: raw1394: arm functions slept in atomic context
ieee1394: sbp2: enable auto spin-up for all SBP-2 devices
MAINTAINERS: updates to IEEE 1394 subsystem maintainership
ieee1394: ohci1394: check for errors in suspend or resume
set power state of firewire host during suspend
ieee1394: ohci1394: more obvious endianess handling
ieee1394: ohci1394: fix endianess bug in debug message
ieee1394: sbp2: don't prefer MODE SENSE 10
ieee1394: nodemgr: grab class.subsys.rwsem in nodemgr_resume_ne
ieee1394: nodemgr: fix rwsem recursion
ieee1394: sbp2: more help in Kconfig
ieee1394: sbp2: prevent rare deadlock in shutdown
ieee1394: sbp2: update includes
ieee1394: sbp2: better handling of transport errors
ieee1394: sbp2: recheck node generation in sbp2_update
ieee1394: sbp2: safer agent reset in error handlers
ieee1394: sbp2: handle "sbp2util_node_write_no_wait failed"
CONFIG_PM=n slim: drivers/ieee1394/ohci1394.c
ieee1394: safer definition of empty macros
video1394: add poll file operation support
...
i2c-algo-bit: Discard the mdelay data struct member
The i2c_algo_bit_data structure has an mdelay member, which is not
used by the algorithm code (the code has always been ifdef'd out.)
Let's discard it to save some code and memory.
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Mauro Carvalho Chehab <mchehab@brturbo.com.br>
Cc: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Sleeping functions like copy_to_user were accessed inside spinlocks in
raw1394's arm_register, arm_unregister, arm_get_buf, arm_set_buf.
http://bugzilla.kernel.org/show_bug.cgi?id=7120
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Tested-by: David Trent <DTrent@piacton.com>
(cherry picked from e575953ec17c3f5c1e738847d2d16c241bb99783 commit)
This is a follow-up to patch "ieee1394: sbp2: enable auto spin-up for
Maxtor disks". When I 'ejected' an OXUF922 based HDD from a Mac OS X
box, it was spun down by the Mac and did not spin up by itself when
attached to a Linux box right after that. The first SCSI command that
required the bridge to access the drive ended in
sda:<6>sd 18:0:0:0: Device not ready: <6>: Current: sense key: Not Ready
Additional sense: Logical unit not ready, initializing cmd. required
Therefore the flag which instructs scsi_mod to send START STOP UNIT with
START=1 ("make medium ready") after such a condition is now enabled
unconditionally for all FireWire storage devices.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Put firewire host controller in PCI Dx state for system suspend.
(I was not able to measure any power savings, but it sounds like right
thing to do, anyway.)
Signed-off-by: Pavel Machek <pavel@suse.cz>
Update by stefanr: Shuffle with existing PPC_PMAC code. Set power
state in the resume hook too.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Rename ohci1394's packet_swab to header_le32_to_cpu to better reflect
what it actually does. Also, define a constant array as 'const' and
check the array index properly.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The transaction labels were misprinted int the debug printk "Packet
received from node..." due two byte-swapping once too often. Affected
were big endian machines, except UniNorth based ones. Fix tested by
Wolfgang Pfeiffer.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
In the old days, sbp2 used to coerce all MODE SENSE commands into the
10 bytes version. When all command set conversions were removed from
sbp2 several months ago, sdev->use_10_for_ms = 1 was added. Meaning,
higher SCSI layers preferred the 10 bytes version but would try the 6
bytes version if the former failed.
Recently, a problem with the 10 bytes version was discovered. An Initio
INIC-1530 firmware accepted the 10 bytes version but replied with bogus
data, showing the HDD incorrectly as write-protected. Since RBC
actually mandates MODE SENSE (6), I checked which version was sent by
Windows XP and Mac OS X 10.3 to an SBP-2 target hosted by Linux --- it
was the 6 bytes version. (Exception: OS X sent the 10 bytes version to
an MMC target. RBC and SBC got MODE SENSE (6).)
Therefore, drop the use_10_for_ms flag from sbp2. Now the upper layers
will try MODE SENSE (6) before MODE SENSE (10) on all SBP-2 devices.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
nodemgr_resume_ne was iterating over nodemgr_ud_class.children without
protection by nodemgr_ud_class.subsys.rwsem.
FIXME:
Shouldn't we rather use class->sem there, not class->subsys.rwsem?
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
nodemgr_update_pdrv grabbed an rw semaphore (as reader) which was
already taken by its caller's caller, nodemgr_probe_ne (as reader too).
Reported by Miles Lane, call path pointed out by Arjan van de Ven.
FIXME:
Shouldn't we rather use class->sem there, not class->subsys.rwsem?
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Scsi_remove_device() may go into uninterruptible sleep if blocked.
Therefore sbp2_remove() unblocks the Scsi_Host before the device is
requested to be removed. But there could be another 1394 bus reset
after that which would block the host again. The 1394 subsystem won't
call sbp2_update() concurrently to sbp2_remove(), which is why there is
no chance for sbp2_remove() to be unblocked by sbp2_update().
The fix is to tell sbp2's bus reset handler when a device is to be shut
down so that it skips scsi_block_requests() on that host. As before,
any new commands after a reset without reconnect will be failed quickly
by sbp2scsi_queuecommand().
In the long term, means to go without scsi_block_requests() should be
found.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Remove unused includes. Add missing includes, i.e. explicitly include
all used headers. Sort includes alphabetically. Replace one call to
signal_pending(current) to avoid to include headers just for this line.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
If the target signals a transport failure via status block, complete the
request with DID_BUSY to indicate to the SCSI subsystem that the command
may succeed when retried.
Also log diagnostic information if the status block shows a transport
related problem. It may point to hardware faults.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
While sbp2_update() is doing its duties after a bus reset, another reset
could happen. Don't accept new requests until the next undisturbed
sbp2_update() or until sbp2_remove().
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The scsi_host_template's eh_abort_handler and eh_device_reset_handler
are allowed to sleep. Use this to run sbp2_agent_reset in the more
reliable mode which returns _after_ its write transaction was finished.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Fix for http://bugzilla.kernel.org/show_bug.cgi?id=6948
Because sbp2 writes to the target's fetch agent's registers from within
atomic context, it cannot sleep to guaranteedly get a free transaction
label. This may repeatedly lead to "sbp2util_node_write_no_wait failed"
and consequently to SCSI command abortion after timeout. A likely cause
is that many queue_command softirqs may occur before khpsbpkt (the
ieee1394 driver's thread which cleans up after finished transactions) is
woken up to recycle tlabels.
Sbp2 now schedules a workqueue job whenever sbp2_link_orb_command fails
in sbp2util_node_write_no_wait. The job will reliably get a transaction
label because it can sleep.
We use the kernel-wide shared workqueue because it is unlikely that the
job itself actually needs to sleep. In the improbable case that it has
to sleep, it doesn't need to sleep long since the standard transaction
timeout is 100ms.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Remove some code which is unneeded if CONFIG_PM=n.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
A deactivated macro, defined as "#define foo(bar)", will result in
silent corruption if somebody forgets a semicolon after a call to foo.
Replace it by "#define foo(bar) do {} while (0)" which will reveal any
respective syntax errors.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This patch adds support for the poll file operation to the video1394
driver.
Signed-off-by: David Moore <dcm@acm.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This patch contains the scheduled removal of the force_inquiry_hack
module parameter.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This makes debugging with firescope easier.
Signed-off-by: Andi Kleen <ak@suse.de> (original patch)
Update:
- no need for #ifdef MODULE
- add comment in ieee1394_core, more verbose comment in ohci1394
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de> (update)
The waitqueue API is used to replace a custom wait mechanism. Only one
global waitqueue (instead of per-device waitqueues or completions) is
added because there is usually just one waiter.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
- Add checks for the (very unlikely) cases that the target writes too
little or too much status data or writes unsolicited status.
- Indicate that these and similar conditions are unlikely().
- Check the 'resp' and 'sbp_status' fields for possible failure status.
- Slightly optimize access macros for the status block bitfields.
- Unify a few related log messages.
TODO: Check if 'src'==1, then withhold the respective ORB from reuse
until status for any subsequent ORB was received. This is an old bug
whose fix requires more complex command queue handling.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Sbp2's copy of the status fifo was cleared when management ORBs or new
command ORBs were prepared. The latter had potential for a race
condition if the block layer's soft IRQ and the 1394 LLD's interrupt
handler ran on different CPUs. It would also yield wrong status if a
command was completed with non-zero completion status before other
commands that had zero completion status, and no new command was
enqueued in the meantime.
Now, the status buffer is cleared right before it is written. Thus it
ends up in the following simpler and safer access pattern:
- sbp2_alloc_device: allocates and implicitly clears once,
- sbp2_handle_status_write: clears, writes, and reads,
- sbp2_query_logins, sbp2_login_device, sbp2_reconnect_device: read.
The latter three do not race with sbp2_handle_status_write because of
how the protocol works.
As a tiny optimization, the first two quadlets of the status never need
to be cleared.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Only the driver writes ORBs, the device just reads them. Therefore
PCI_DMA_BIDIRECTIONAL can be replaced by PCI_DMA_TODEVICE which may be
cheaper on some architectures.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Since sbp2 is at the moment unable to do anything with the return value
of sbp2_link_orb_command, just discard it.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The sbp2 initiator has two ways to tell a target's fetch agent about new
command ORBs:
- Write the ORB's address to the ORB_POINTER register. This must not
be done while the fetch agent is active.
- Put the ORB's address into the previously submitted ORB's next_ORB
field and write to the DOORBELL register. This may be done while the
fetch agent is active or suspended. It must not be done while the
fetch agent is in reset state.
Sbp2 has a last_orb pointer which indicates in what way a new command
should be announced. That pointer is concurrently accessed at various
occasions. Furthermore, initiator and target are accessing the next_ORB
field of ORBs concurrently and asynchronously.
This patch does:
- Protect all initiator accesses to last_orb by sbp2_command_orb_lock.
- Add pci_dma_sync_single_for_device before a previously submitted
ORB's next_ORB field is overwritten.
- Insert a memory barrier between when next_ORB_lo and next_ORB_hi are
overwritten. Next_ORB_hi must not be updated before next_ORB_lo.
- Remove the rather unspecific and now superfluous qualifier "volatile"
from the next_ORB fields.
- Add comments on how last_orb is connected with what is known about
the target's fetch agent's state.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This patch reduces the size of struct hpsb_host and also removes
semaphores from ieee1394_transactions.c. On i386, struct hpsb_host
shrinks from 10656 bytes to 6688 bytes. This is accomplished by
- using a single wait_queue for hpsb_get_tlabel instead of many
instances of semaphores,
- using a single lock to serialize access to all tlabel pools (the
protected code regions are small, i.e. lock contention very low),
- omitting the sysfs attribute tlabels_allocations.
Drawback: In the rare case that a process needs to sleep because all
transaction labels for the node are temporarily exhausted, it is also
woken up if a tlabel for a different node became free, checks for an
available tlabel, and is put to sleep again. The check is not costly
and the situation occurs extremely rarely. (Tlabels are typically
only exhausted if there was no context switch to the khpsbpkt thread
which recycles tlables.) Therefore the benefit of reduced tpool size
outweighs this drawback.
The sysfs attributes tlabels_free and tlabels_mask are not compiled
anymore unless CONFIG_IEEE1394_VERBOSEDEBUG is set.
The by far biggest member of struct hpsb_host, the struct csr_control
csr (5272 bytes on i386), is now placed at the end of struct hpsb_host.
Note, hpsb_get_tlabel calls the macro wait_event_interruptible with a
condition argument which has a side effect (allocation of a tlabel and
manipulation of the packet). This side effect happens only if the
condition is true. The patch relies on wait_event_interruptible not
evaluating the condition again after it became true.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Recently a patch was added for preliminary suspend/resume handling on
!PPC_PMAC. However, this broke both suspend and firewire on powerpc
because it saves the pci state after the device has already been disabled.
This moves the save state to before the pmac specific code.
Signed-off-by: Danny Tholen <obiwan@mailmij.org>
Cc: Stefan Richter <stefanr@s5r6.in-berlin.de>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Ben Collins <bcollins@ubuntu.com>
Cc: Jody McIntyre <scjody@modernduck.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
At least Maxtor OneTouch III require a "start stop unit" command after auto
spin-down before the next access can proceed. This patch activates the
responsible code in scsi_mod for all Maxtor SBP-2 disks.
https://bugzilla.novell.com/show_bug.cgi?id=183011
Maybe that should be done for all SBP-2 disks, but better be cautious.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Cc: Jody McIntyre <scjody@modernduck.com>
Cc: Ben Collins <bcollins@ubuntu.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
ieee1394 reuses the skb infrastructure of the networking code, and uses two
skb-head queues: ->pending_packet_queue and hpsbpkt_queue. The latter is used
in the usual fashion: processed from a kernel thread. The other one,
->pending_packet_queue is also processed from hardirq context (f.e. in
hpsb_bus_reset()), which is not what the networking code usually does (which
completes from softirq or process context). This locking assymetry can be
totally correct if done carefully, but it can also be dangerous if networking
helper functions are reused, which could assume traditional networking use.
It would probably be more robust to push this completion into a workqueue -
but technically the code can be 100% correct, and lockdep has to be taught
about it. The solution is to split the ->pending_packet_queue skb-head->lock
class from the networking lock-class by using a private lock-validator key.
Has no effect on non-lockdep kernels.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Stefan Richter <stefanr@s5r6.in-berlin.de>
Cc: Jody McIntyre <scjody@modernduck.com>
Cc: Ben Collins <bcollins@debian.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There was stuff between the comment and the function.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
Another trivial sem2mutex conversion.
Side note: nodemgr_serialize's purpose, when introduced in linux1394's
revision 529 in July 2002, was to protect several data structures which
are now largely handled by or together with Linux' driver core and are
now protected by the LDM's own mechanisms. It may very well be possible
to remove this mutex now. But fully parallelized node scanning is on
our long-term TODO list anyway; the mutex will certainly go away then.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
Convert nodemgr's host thread from kernel_thread to kthread and its
sleep/restart mechanism from a counting semaphore to a schedule()/
wake_up_process() scheme.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
Nodemgr's ignore_drivers variable is exposed as a module load parameter
(therefore also as a sysfs attribute below /sys/module) and additionally
as an attribute below /sys/bus/ieee1394. Since the latter is writable,
make the former writable too.
Note, the bus's attribute ignore_drivers is only relevant to newly added
units, not to present or suspended or resuming units. Those have their
own attribute ignore_driver.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
nodemgr.c::fw_set_rescan() is used to re-run the driver core over
nodemgr's representation of unit directories in order to initiate
protocol driver probes. It is initiated via write access to one of
nodemgr's sysfs attributes. The purpose is to attach drivers to
units after switching a unit's ignore_driver attribute from 1 to 0.
It is not really necessary to fork a kernel_thread for this job. The
call to kernel_thread() can be eliminated to avoid the deprecated API
and to simplify the code a bit.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
An already existing wait queue replaces raw1394's complete_sem which was
maintained in parallel to the wait queue. The role of the semaphore's
counter is taken over by a direct check of what was really counted: The
presence of items in the list of completed requests.
Notes:
- raw1394_release() sleeps uninterruptibly until all requests were
completed. This is the same behaviour as before the patch.
- The macros wait_event and wait_event_interruptible are called with a
condition argument which has a side effect, i.e. manipulation of the
requests list. This side effect happens only if the condition is
true. The patch relies on the fact that wait_event[_interruptible]
does not evaluate the condition again after it became true.
- The diffstat looks unfavorable with respect to added lines of code.
However 19 of them are comments, and some are due to separation of
existing code blocks into two small helper functions.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
hpsb_update_config_rom() is defined in csr.c, not hosts.c.
hpsb_get_config_rom() does not exist.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
Remove unnecessary includes, add missing includes.
Use forward type declarations for some structs.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
Adjust tabulators, line wraps, empty lines, and comment style.
Update comments in ieee1394_transactions.h and highlevel.h.
Fix typo in comment in csr.h.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Ben Collins <bcollins@ubuntu.com>
