Pull firewire updates from Takashi Sakamoto:
"This update includes the following changes:
- Removal of the deprecated debug parameter from firewire-ohci module
- Replacement of the module-local workqueue in 1394 OHCI PCI driver
with a companion IRQ thread
- Refactoring of bus management code
- Additional minor code cleanup
The existing tracepoints serve as an alternative to the removed debug
parameter. The use of IRQ thread is experimental, as it handles 1394
OHCI SelfIDComplete event only. It may be replaced in the future
releases with another approach; e.g. by providing workqueue from core
functionality"
* tag 'firewire-updates-6.18' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394: (43 commits)
firewire: core: fix undefined reference error in ARM EABI
Revert "firewire: core: disable bus management work temporarily during updating topology"
Revert "firewire: core: shrink critical section of fw_card spinlock in bm_work"
firewire: core: suppress overflow warning when computing jiffies from isochronous cycle
firewire: core: minor code refactoring to delete useless local variable
firewire: core; eliminate pick_me goto label
firewire: core: code refactoring to split contention procedure for bus manager
firewire: core: code refactoring for the case of generation mismatch
firewire: core: use switch statement to evaluate transaction result to CSR_BUS_MANAGER_ID
firewire: core: remove useless generation check
firewire: core: use struct_size and flex_array_size in ioctl_add_descriptor
firewire: core: shrink critical section of fw_card spinlock in bm_work
firewire: core: disable bus management work temporarily during updating topology
firewire: core: schedule bm_work item outside of spin lock
firewire: core: annotate fw_destroy_nodes with must-hold-lock
firewire: core: use spin lock specific to timer for split transaction
firewire: core: use spin lock specific to transaction
firewire: core: use spin lock specific to topology map
firewire: core: maintain phy packet receivers locally in cdev layer
firewire: core: use scoped_guard() to manage critical section to update topology
...
In kernel v6.5, several functions were added to the cdev layer. This
required updating the default version of subsystem ABI up to 6, but
this requirement was overlooked.
This commit updates the version accordingly.
Fixes: 6add87e976 ("firewire: cdev: add new version of ABI to notify time stamp at request/response subaction of transaction#")
Link: https://lore.kernel.org/r/20250920025148.163402-1-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Current implementation stores the bus generation at which the bus manager
contending procedure finishes. The condition for the procedure is the
mismatch of the stored generation against current bus generation.
This commit refactors the code for the contending procedure. Two existing
branches are put into a new branch to detect the generation mismatch, thus
the most of change is indentation.
Link: https://lore.kernel.org/r/20250918235448.129705-4-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Two functions, fw_core_handle_bus_reset() and bm_work(), are serialized
by a commit 3d91fd440cc7 ("firewire: core: disable bus management work
temporarily during updating topology"). Therefore the generation member
of fw_card is immutable in bm_work().
This commit removes useless generation check in bm_work().
Link: https://lore.kernel.org/r/20250918235448.129705-2-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
When processing selfID sequence, bus topology tree is (re)built, and some
members of fw_card are determined. Once determined, the members are valid
during the bus generation. The above operations are in the critical
section of fw_card spin lock.
Before building the bus topology, a work item is scheduled for bus manager
work. The bm_work() function is invoked by the work item. The function
tries to acquire the spin lock, then can be stalled until the bus topology
building finishes.
The bus manager should work once the members of fw_card are determined.
This commit suppresses the above situation by disabling the work item
during processing selfID sequence.
Link: https://lore.kernel.org/r/20250917000347.52369-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Before (re)building topology tree, fw_core_handle_bus_reset() schedules
a work item under acquiring fw_card spin lock. The work item invokes
bm_work() which acquires the spin lock at first, then can be stalled to
wait until the building tree finishes. This is inconvenient.
This commit moves the timing to schedule the work item after releasing
the spin lock.
Link: https://lore.kernel.org/r/20250917000347.52369-2-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
At present the parameters to compute timeout time for split transaction is
protected by card-wide spin lock, while it is not necessarily convenient
in a point to narrower critical section.
This commit adds and uses another spin lock specific for the purpose.
Link: https://lore.kernel.org/r/20250915234747.915922-6-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The list of instance for asynchronous transaction to wait for response
subaction is maintained as a member of fw_card structure. The card-wide
spinlock is used at present for any operation over the list, however it
is not necessarily suited for the purpose.
This commit adds and uses the spin lock specific to maintain the list.
Link: https://lore.kernel.org/r/20250915234747.915922-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
At present, the operation for read transaction to topology map register is
not protected by any kind of lock primitives. This causes a potential
problem to result in the mixed content of topology map.
This commit adds and uses spin lock specific to topology map.
Link: https://lore.kernel.org/r/20250915234747.915922-4-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The value of BUS_MANAGER_ID register has 0x3f when no node_id is
registered. Current implementation uses hard-coded numeric literal but
in the case the macro expression is preferable since it is easy to
distinguish the state from node ID mask.
This commit applies the idea.
Link: https://lore.kernel.org/r/20250913105737.778038-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The bm_work work item should be scheduled after holding fw_card reference
counting. At a commit 25feb1a96e ("firewire: core: use cleanup function
in bm_work"), I misinterpreted it as fw_card spinlock and inserted
lockdep_assert_hold() wrongly.
This commit removes the useless line.
Fixes: 25feb1a96e ("firewire: core: use cleanup function in bm_work")
Link: https://lore.kernel.org/r/20250911221312.678076-1-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
In the middle of bm_work function, both the value of gap_count and the
state of root node are investigated. Current implementation is not a good
shape since the investigation is aligned to be flat.
This commit refactors the investigation with two large branches.
Link: https://lore.kernel.org/r/20250908012108.514698-11-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The call of bm_work should be done after acquiring spin lock of fw_card.
For asynchronous transaction, the lock should be released temporarily
due to event waiting.
A commit 27310d5616 ("firewire: core: use guard macro to maintain
properties of fw_card") applied scoped_guard() to the bm_work function,
however it looks hard to follow to the control flow.
This commit refactors the spin lock acquisition after the transaction.
Link: https://lore.kernel.org/r/20250908012108.514698-8-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The data mbmer in fw_node structure is an opaque pointer, while nowadays
it is just used to refer to fw_device associated with the fw_node.
This commit redefines the opaque pointer to a pointer to fw_device
structure, and adds some helper functions to set/get it.
Link: https://lore.kernel.org/r/20250908012108.514698-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The first step maintaining the bus topology is to handle SelfIDComplete
event. This event occurs after initiating bus reset when 1394 OHCI link
layer is enabled, or when the bus topology changes (e.g. when a device is
added). Because enumeration of the selfID sequence can take some time, it
should be processed in a bottom half.
Currently, this is done in a module-local workqueue with the
WQ_MEM_RECLAIM flag, to allow invocation during memory reclaim paths. A
threaded IRQ handler is a preferable alternative, as it eliminates the
need to manage workqueue attributes manually.
Although SelfIDComplete events are not so frequent in normal usage,
handling them correctly is critical for proper bus topology management.
This commit switches SelfIDComplete handling to a threaded IRQ handler.
Link: https://lore.kernel.org/r/20250823030954.268412-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Between v6.11 and v6.12, a set of tracepoints was added to record
asynchronous communication events:
- firewire:async_phy_inbound
- firewire:async_phy_outbound_initiate
- firewire:async_phy_outbound_complete
- firewire:async_response_inbound
- firewire:async_response_outbound_initiate
- firewire:async_response_outbound_complete
- firewire:async_request_inbound
- firewire:async_request_outbound_initiate
- firewire:async_request_outbound_complete
These tracepoints cover the functionality of the existing debug logging.
This commit removes the logging.
Link: https://lore.kernel.org/r/20250821003017.186752-4-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
A commit 677ceae190 ("firewire: core: add tracepoints event for
self_id_sequence") added the "firewire:self_id_sequence" event in v6.11.
A commit 526e21a2aa ("firewire: ohci: add tracepoints event for data
of Self-ID DMA") added the "firewire_ohci:self_id_complete" event in
v6.12.
These tracepoints replace the equivalent debug logging. This commit
removes the logging.
Link: https://lore.kernel.org/r/20250821003017.186752-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The former commit has a limitation that only up to 4 FCP address
handlers could be processed per request. Although it suffices for most
use cases, it is technically a regression.
This commit lifts the restriction by reallocating the buffer from kernel
heap when more than 4 handlers are registered. The allocation is performed
within RCU read-side critical section, thus it uses GCP_ATOMIC flag. The
buffer size is rounded up to the next power of two to align with kmalloc
allocation units.
Link: https://lore.kernel.org/r/20250803122015.236493-5-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The former commit added reference counting to ensure safe invocations of
address handlers. Unlike the exclusive-region address handlers, all FCP
address handlers should be called on receiving an FCP request.
This commit uses the part of kernel stack to collect address handlers up
to 4 within the section, then invoke them outside of the section.
Reference counting ensures that each handler remains valid and safe to
call.
Lifting the limitation of supporting only 4 handlers is left for next
work.
Link: https://lore.kernel.org/r/20250803122015.236493-4-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The previous commit added reference counting to ensure safe invocations of
address handlers.
This commit moves the invocation of handlers for exclusive regions outside
of the RCU read-side critical section. The address handler for the
requested region is selected within the critical section, then invoked
outside of it.
Link: https://lore.kernel.org/r/20250803122015.236493-3-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
The lifetime of address handler has been managed by linked list and RCU.
This approach was introduced in commit 35202f7d84 ("firewire: remove
global lock around address handlers, convert to RCU"). The invocations of
address handler are performed within RCU read-side critical sections.
In commit 57e6d9f85f ("firewire: ohci: use workqueue to handle events
of AR request/response contexts"), the invocations are in a workqueue
context. The approach still imposes limitation that sleeping is not
allowed within RCU read-side critical sections. However, since sleeping
is not permitted within RCU read-side critical sections, this approach
still has a limitation.
This commit adds reference counting to decouple handler invocation from
handler discovery. The linked list and RCU is used to discover the
handlers, while the reference counting is used to invoke them safely.
Link: https://lore.kernel.org/r/20250803122015.236493-2-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
In commit 386a4153a2 ("firewire: ohci: cache the context run bit"), a
running member was added to the context structure to cache the running
state of a given DMA context. Although this member is accessible from IR,
IT, and AT contexts, it is currently used only by the AT context.
Additionally, the context structure includes a work item, which is also
used by the AT context. Both members are unnecessary for IR and IT
contexts.
This commit refactors the code by moving these two members into a new
structure specific to AT context.
Link: https://lore.kernel.org/r/20250710131916.31289-1-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
