A compiler barrier (explicit on the read side, implicit on the write
side) is not quite enough for what has to be accomplished here. Use
hardware memory barriers on systems which need them.
(Of course a full fix of generation handling would require much more
than this. The ieee1394 core's bus generation counter had to be tied to
the controller's bus generation counter; cf. Kristian's stack. It's
just that I have other current business with the code around these
barrier()s, so why not do at least this small fix.)
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
It is already known that buggy firmwares exist which report a bogus
link_spd in their config ROM bus info block. We now got the first
report of a bogus max_rom too (Freecom FireWire Hard Drive 1TB,
http://bugzilla.kernel.org/show_bug.cgi?id=12206).
I suspect other OSs only use quadlet reads to fetch the config ROM,
otherwise the firmware authors would have noticed their mistake.
Hence limit ieee1394's config ROM fetching routine to quadlets as the
safe minimum regardless of what the bus info block says.
This will potentially slow the bus reset handling by nodemgr somewhat
down. But most existing devices support only quadlet reads anyway,
hence there will often be no actual difference to before this change.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
According to http://bugzilla.kernel.org/show_bug.cgi?id=12206, Freecom
FireWire Hard Drive 1TB reports max_rom=2 but returns garbage if block
read requests are used to read the config ROM. Force max_rom=0 to limit
them to quadlet read requests.
Reported-by: Christian Mueller <cm1@mumac.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The firewire nodemanager function "nodemgr_host_thread" contains a loop
that calls try_to_freeze near the top of the loop, but then delays for
up to 3.25 seconds (plus time to do work) before getting back to the top
of the loop. When starting a cycle post-boot, this doesn't seem to bite,
but it is causing a noticeable delay at boot time, when freezing
processes prior to starting to read the image.
The following patch adds invocation of try_to_freeze to the subloops
that are used in the body of this function. With these additions, the
time to freeze when starting to resume at boot time is virtually zero.
I'm no expert on firewire, and so don't know that we shouldn't check
the return value and jump back to the top of the loop or such like after
being frozen, but I submit it for your consideration.
Signed-off-by: Nigel Cunningham <nigel@tuxonice.net>
The delay until nodemgr freezes was up to 0.25s (plus time for node
probes) in Linux 2.6.27 and older and up to 3.25s (plus ~) since Linux
2.6.28-rc1, hence much more noticeable.
try_to_freeze() without any jump is correct. The surrounding code in
the respective loops will catch whether another bus reset happens during
the freeze and handle it.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
There are situations when nodes vanish from the bus and come back in
quickly thereafter:
- When certain bus-powered hubs are plugged in,
- when certain disk enclosures are switched from self-power to bus
power or vice versa and break the daisy chain during the transition,
- when the user plugs a cable out and quickly plugs it back in, e.g.
to reorder a daisy chain (works on Mac OS X if done quickly enough),
- when certain hubs temporarily malfunction during high bus traffic.
The ieee1394 driver's nodemgr already contained a function to set
vanished nodes aside into "limbo"; i.e. they wouldn't actually be
deleted right away. (In fact, only unloading the driver or writing into
an obscure sysfs attribute would delete them eventually.) If nodes
reappeared later, they would be resurrected out of limbo.
Moving nodes into and out of limbo was accompanied with calling the
.suspend() and .resume() driver methods of the drivers which were bound
to a respective node's unit directories. Not only is this somewhat
strange due to the intended use of these driver methods for power
management, also the sbp2 driver in particular does not implement
.suspend() and .resume(). Hence sbp2 would be disconnected from devices
in situations as listed above.
We now:
- leave drivers bound when nodes go into limbo,
- call the drivers' .update() when nodes come out of limbo,
- automatically delete in-limbo nodes 3 seconds after the last
bus reset and bus rescan.
- Because of the automatic removal, the now obsolete bus attribute
/sys/bus/ieee1394/destroy_node is removed.
This especially lets sbp2 survive brief disconnections. You can for
example yank a disk's cable and plug it back in while reading the
respective disk with dd, but dd will happily continue as if nothing
happened.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Remove useless pointer type casts.
Remove unnecessary hi->host indirection where only host is used.
Remove an unnecessary WARN_ON.
Change a few names.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
nodemgr_node_probe checked for generation increments too late and
therefore prematurely reported nodes as "suspended".
Fixes http://bugzilla.kernel.org/show_bug.cgi?id=11349. Reported and
tested by Damien Benoist.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Regression since commit 73cf60232e,
"ieee1394: use class iteration api": The two loops for (1.) driver
updates and (2.) driver probes were replaced by a single loop with
bogus needs_probe checks. Hence updates and probes were now intermixed,
and especially sbp2 updates (reconnects) held up longer than necessary.
While we fix it, change the needs_probe flag to bool type for clarity.
Tested by Damien Benoist.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This mirrors the functionality that driver_find_device has as well.
We add a start variable, and all callers of the function are fixed up at
the same time.
The block layer will be using this new functionality in a follow-on
patch.
Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This mirrors the functionality that driver_for_each_device has as well.
We add a start variable, and all callers of the function are fixed up at
the same time.
The block layer will be using this new functionality in a follow-on
patch.
Cc: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
firewire: fw-sbp2: log scsi_target ID at release
ieee1394: fix NULL pointer dereference in sysfs access
Regression since "ieee1394: prevent device binding of raw1394,
video1394, dv1394", commit d2ace29fa44589da51fedc06a67b3f05301f3bfd:
$ cat /sys/bus/ieee1394/drivers/raw1394/device_ids
triggers a NULL pointer dereference in fw_show_drv_device_ids.
Reported by Miles Lane.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Tested-by: Miles Lane <miles.lane@gmail.com>
These drivers don't need to match any unit_directory type device.
They just need the id_table for module autoloading per module alias.
Not binding any of these drivers allows special-purpose drivers with
similar or same IDs to bind to devices. This currently only benefits
out-of-tree drivers; on the other hand it is in no way detrimental to
in-tree drivers.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Convert to use the class iteration api.
Signed-off-by: Dave Young <hidave.darkstar@gmail.com>
Cc: Stefan Richter <stefanr@s5r6.in-berlin.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
At least since nodemgr got rid of coarse global locking, accesses to
struct csr1212_keyval's reference counter should be atomic and coupled
with proper barriers. Also, calls to csr1212_keep_keyval(kv) should
occur before kv is being used.
(We probably should convert refcnt to struct kref, but how to keep
csr1212_destroy_keyval's implementation non-recursively then?)
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
csr1212_keep_keyval(kv) in nodemgr_process_root_directory was
unbalanced if ne->vendor_name_kv already exists. This happens for
example if eth1394 or raw1394 modify the local config ROM and it is
parsed again.
As a bonus, the attempt to add the vendor_name_kv sysfs attribute
when it already exists is now fixed for good.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The nodemgr host thread can exit on its own even when kthread_should_stop
is not true, on receiving a signal (might never happen in practice, as
it ignores signals). But considering kthread_stop() must not be mixed with
kthreads that can exit on their own, I think changing the code like this
is clearer. This change means the thread can cut its sleep short when
receive a signal but looking at the code around, that sounds okay (and
again, it might never actually recieve a signal in practice).
Signed-off-by: Satyam Sharma <satyam@infradead.org>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This changes the uevent buffer functions to use a struct instead of a
long list of parameters. It does no longer require the caller to do the
proper buffer termination and size accounting, which is currently wrong
in some places. It fixes a known bug where parts of the uevent
environment are overwritten because of wrong index calculations.
Many thanks to Mathieu Desnoyers for finding bugs and improving the
error handling.
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Cc: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently, the freezer treats all tasks as freezable, except for the kernel
threads that explicitly set the PF_NOFREEZE flag for themselves. This
approach is problematic, since it requires every kernel thread to either
set PF_NOFREEZE explicitly, or call try_to_freeze(), even if it doesn't
care for the freezing of tasks at all.
It seems better to only require the kernel threads that want to or need to
be frozen to use some freezer-related code and to remove any
freezer-related code from the other (nonfreezable) kernel threads, which is
done in this patch.
The patch causes all kernel threads to be nonfreezable by default (ie. to
have PF_NOFREEZE set by default) and introduces the set_freezable()
function that should be called by the freezable kernel threads in order to
unset PF_NOFREEZE. It also makes all of the currently freezable kernel
threads call set_freezable(), so it shouldn't cause any (intentional)
change of behaviour to appear. Additionally, it updates documentation to
describe the freezing of tasks more accurately.
[akpm@linux-foundation.org: build fixes]
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Nigel Cunningham <nigel@nigel.suspend2.net>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Gautham R Shenoy <ego@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Going through the string and waiting for _pointer_ to become '\0'
is not what the authors meant...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Ben Collins <ben.collins@ubuntu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove the global nodemgr_serialize mutex which enclosed most of the
host thread event loop. This allows for parallelism between several
host adapter cards.
Properly serialize the driver hooks .update(), .suspend(), .resume(),
and .remove() by means of device->sem. These hooks can be called from
outside the host threads' contexts.
Get() and put() the device.driver when calling its hooks.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Here is a straightforward conversion to "struct device". The "struct
class_device" will be removed from the kernel.
It seems to work fine for me with and without CONFIG_SYSFS_DEPRECATED
set.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
With "modprobe sbp2 long_ieee1394_id=y", the format of
/sys/bus/scsi/devices/*:*:*:*/ieee1394_id is changed from e.g.
0001041010004beb:0:0 to 0001041010004beb:00042c:0000.
The longer format fully conforms to object identifier sizes as per
SAM(-2...4) and reflects what the SAM target port identifier is meant to
contain: A Discovery ID allegedly specified by ISO/IEC 13213:1994 ---
however there is no such thing; the authors of SAM probably meant
Directory ID). Especially target nodes with multiple dynamically added
targets may use Directory IDs to persistently identify target ports.
The new format is independent of implementation details of nodemgr.
Thus the same ieee1394_id attribute format can be implemented in the new
firewire stack.
The ieee1394_id is typically used to create persistently named links in
/dev/disk/by-id.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Everytime when eth1394 or a libraw1394 client updates the configuration
ROM, a certain sysfs attribute cannot be added since it already exists.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
csr1212 was written to be compiled either as part of the ieee1394 kernel
driver or of an anticipated IEEE 1212 userspace library. We now drop
support for the latter. The costs in terms of code footprint and depth
of abstraction are not countered by any actual benefit.
Also remove some obsolete #includes.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
The subsystem rwsem is not used by the driver core at all, so the use of
it in the ieee1394 code doesn't make any sense. They might possibly
want to use a local lock, but as most of these operations are already
protected by a local lock, it really doesn't look like it would be
needed.
Cc: Ben Collins <bcollins@debian.org>
Cc: Stefan Richter <stefanr@s5r6.in-berlin.de>
Cc: linux1394-devel <linux1394-devel@lists.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Make use of add_uevent_var() instead of (often incorrectly) open coding it.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au>
Signed-off-by: Eric Rannaud <eric.rannaud@gmail.com>
Cc: Kay Sievers <kay.sievers@vrfy.org>
Cc: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
A "modprobe ohci1394; sleep 1.5; modprobe -r ohci1394" could get stuck
in uninterruptible state, especially if an external node was connected.
http://bugzilla.kernel.org/show_bug.cgi?id=7792
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Since my commit 8252bbb136 in 2.6.20-rc1,
host devices have a dummy driver attached. Alas the driver was not
registered before use if ieee1394 was loaded with disable_nodemgr=1.
This resulted in non-functional FireWire drivers or kernel lockup.
http://bugzilla.kernel.org/show_bug.cgi?id=7942
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This patch contains the scheduled IEEE1394_OUI_DB removal.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Update: Also remove drivers/ieee1394/.gitignore.
Remove now unused struct members in drivers/ieee1394/nodemgr.h.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
This patch consolidates some bookkeeping for driver registering. It
closely models what pci_register_driver() does. The main addition is
that the owner of the driver is set, so we get a proper symlink
for /sys/bus/ieee1394/driver/*/module.
Also moves setting of name and bus type into nodemgr. Because of this,
we can remove the EXPORT_SYMBOL for ieee1394_bus_type, since it's now
only used in ieee1394.ko.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
If "modprobe ohci1394" was quickly followed by "modprobe -r ohci1394",
say with 1 second pause in between, the modprobe -r got stuck in
uninterruptible sleep in kthread_stop. At the same time the knodemgrd
slept uninterruptibly in bus_rescan_devices_helper. That's because
driver_detach took the semaphore of the PCI device and
bus_rescan_devices_helper wanted to take the semaphore of the FireWire
host device's parent, which is the same semaphore. This was a regression
since Linux 2.6.16, commit bf74ad5bc4,
"Hold the device's parent's lock during probe and remove".
The fix (or workaround) adds a dummy driver to the hpsb_host device. Now
bus_rescan_devices_helper won't scan the host device anymore. This
doesn't hurt since we have no drivers which will bind to these devices
and it is unlikely that there will ever be such a driver. The dummy
driver is befittingly presented as a representation of ieee1394 itself.
Fixes: http://bugzilla.kernel.org/show_bug.cgi?id=6706
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Fix printk format warning:
drivers/ieee1394/nodemgr.c:364: warning: long long unsigned int format, u64 arg (arg 3)
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
- The list "struct class.children" is supposed to be protected by
class.sem, not by class.subsys.rwsem.
- nodemgr_remove_uds() iterated over nodemgr_ud_class.children without
proper protection. This was never observed as a bug since the code
is usually only accessed by knodemgrd. All knodemgrds are currently
globally serialized. But userspace can trigger this code too by
writing to /sys/bus/ieee1394/destroy_node.
- Clean up access to the FireWire bus type's subsys.rwsem: Access it
uniformly via ieee1394_bus_type. Shrink rwsem protected regions
where possible. Expand them where necessary. The latter wasn't a
problem so far because knodemgr is globally serialized.
This should harden the interaction of ieee1394 with sysfs and lay ground
for deserialized operation of multiple knodemgrds and for implementation
of subthreads for parallelized scanning and probing.
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Stefan Richter