aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/firewire
Commit message (Collapse)AuthorAgeFilesLines
* firewire: core: fix to update generation field in topology mapTakashi Sakamoto2025-11-161-1/+2
| | | | | | | | | | | | The generation field of topology map is updated after initialized by zero. The updated value of generation field is always zero, and is against specification. This commit fixes the bug. Fixes: 7d138cb269db ("firewire: core: use spin lock specific to topology map") Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
* firewire: core: Initialize topology_map.lockVille Syrjälä2025-11-141-0/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Lockdep barfs on the new uninitialized spinlock. Initialize it. protip: enable lockdep (CONFIG_PROVE_LOCKING=y) when doing locking changes firewire_ohci 0000:02:01.1: added OHCI v1.10 device as card 0, 4 IR + 4 IT contexts, quirks 0x11 INFO: trying to register non-static key. The code is fine but needs lockdep annotation, or maybe you didn't initialize this object before use? turning off the locking correctness validator. CPU: 0 UID: 0 PID: 1042 Comm: irq/17-firewire Not tainted 6.17.0-rc2-cl-bisect2-00026-g7d138cb269db #136 PREEMPT Hardware name: Dell Inc. Latitude E5400 /0D695C, BIOS A19 06/13/2013 Call Trace: <TASK> dump_stack_lvl+0x6d/0xa0 register_lock_class+0x783/0x790 ? find_held_lock+0x2b/0x80 ? __mod_timer+0x110/0x320 ? __mod_timer+0x110/0x320 __lock_acquire+0x405/0x2600 lock_acquire+0xca/0x2e0 ? fw_core_handle_bus_reset+0x888/0xca0 [firewire_core] ? fw_core_handle_bus_reset+0x878/0xca0 [firewire_core] ? fw_core_handle_bus_reset+0x878/0xca0 [firewire_core] _raw_spin_lock+0x2e/0x40 ? fw_core_handle_bus_reset+0x888/0xca0 [firewire_core] fw_core_handle_bus_reset+0x888/0xca0 [firewire_core] handle_selfid_complete_event+0x35c/0x7a0 [firewire_ohci] ? irq_thread+0x8d/0x280 irq_thread_fn+0x18/0x50 irq_thread+0x15a/0x280 ? irq_check_status_bit+0x100/0x100 ? lockdep_hardirqs_on+0x78/0x100 ? irq_finalize_oneshot.part.0+0xc0/0xc0 ? irq_forced_thread_fn+0x60/0x60 kthread+0x114/0x200 ? kthreads_online_cpu+0x110/0x110 ret_from_fork+0x158/0x1e0 ? kthreads_online_cpu+0x110/0x110 ret_from_fork_asm+0x11/0x20 </TASK> Reported-by: Erhard Furtner <[email protected]> Fixes: 7d138cb269db ("firewire: core: use spin lock specific to topology map") Signed-off-by: Ville Syrjälä <[email protected]> Signed-off-by: Takashi Sakamoto <[email protected]>
* firewire: init_ohci1394_dma: add missing function parameter documentationNirbhay Sharma2025-10-241-0/+10
| | | | | | | | | | | | | | | | | | | | Add missing kernel-doc parameter descriptions for five functions in init_ohci1394_dma.c to fix documentation warnings when building with W=1. This patch addresses the following warnings: - init_ohci1394_wait_for_busresets: missing @ohci description - init_ohci1394_enable_physical_dma: missing @ohci description - init_ohci1394_reset_and_init_dma: missing @ohci description - init_ohci1394_controller: missing @num, @slot, @func descriptions - setup_ohci1394_dma: missing @opt description Tested with GCC 13.2.0 and W=1 flag. All documentation warnings for these functions have been resolved. Signed-off-by: Nirbhay Sharma <[email protected]> Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
* firewire: core: fix __must_hold() annotationTakashi Sakamoto2025-10-241-1/+1
| | | | | | | | | | The variable name passed to __must_hold() annotation is invalid. This commit fixes it. Fixes: 420bd7068cbf ("firewire: core: use spin lock specific to transaction") Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
* Merge tag 'firewire-updates-6.18' of ↵Linus Torvalds2025-10-017-611/+493
|\ | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394 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 ...
| * firewire: core: fix undefined reference error in ARM EABITakashi Sakamoto2025-09-281-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | For ARM EABI, GCC generates a reference to __aeabi_uldivmod when compiling a division of 64-bit integer with 32-bit integer. This function is not available in Linux kernel. In such cases, helper macros are defined in include/linux/math64.h. This commit replaces the division with div_u64(). Fixes: 8ec6a8ec23b9 ("firewire: core: suppress overflow warning when computing jiffies from isochronous cycle") Reported-by: kernel test robot <[email protected]> Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/ Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * Revert "firewire: core: disable bus management work temporarily during ↵Takashi Sakamoto2025-09-241-8/+0
| | | | | | | | | | | | | | | | | | | | | | | | updating topology" This reverts commit abe7159125702c734e851bc0c52b51cd446298a5. The bus manager work item acquires the spin lock of fw_card again, thus no need to serialize it against fw_core_handle_bus_reset(). Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * Revert "firewire: core: shrink critical section of fw_card spinlock in bm_work"Takashi Sakamoto2025-09-241-8/+30
| | | | | | | | | | | | | | | | | | | | | | This reverts commit 582310376d6e9a8d261b682178713cdc4b251af6. The bus manager work has the race condition against fw_destroy_nodes() called by fw_core_remove_card(). The acquition of spin lock of fw_card is left as is again. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: suppress overflow warning when computing jiffies from ↵Takashi Sakamoto2025-09-241-1/+1
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | isochronous cycle The multiplication by USEC_PER_SEC (=1000000L) may trigger an overflow warning with 32 bit storage. In the case of the subsystem the input value ranges between 800 and 16000, thus the result always fits within 32 bit storage. This commit suppresses the warning by using widening conversion to 64 bit storage before multiplication, then using narrowing conversion to 32 bit storage. Reported-by: kernel test robot <[email protected]> Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/ Fixes: 379b870c28c6 ("firewire: core: use helper macros instead of direct access to HZ") Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: minor code refactoring to delete useless local variableTakashi Sakamoto2025-09-181-11/+4
| | | | | | | | | | | | | | The do_reset local variable has less merit. Let's remove it. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core; eliminate pick_me goto labelTakashi Sakamoto2025-09-181-49/+50
| | | | | | | | | | | | | | This commit uses condition statements instead of pick_me goto label. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: code refactoring to split contention procedure for bus managerTakashi Sakamoto2025-09-181-96/+127
| | | | | | | | | | | | | | | | | | | | The precedure to contend for bus manager has much code. It is better to split it into a helper function. This commit refactors in the point. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: code refactoring for the case of generation mismatchTakashi Sakamoto2025-09-181-92/+90
| | | | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use switch statement to evaluate transaction result to ↵Takashi Sakamoto2025-09-181-26/+24
| | | | | | | | | | | | | | | | | | | | | | | | CSR_BUS_MANAGER_ID The result of the lock transaction to swap bus manager on isochronous resource manager looks like an ad-hoc style. It is hard to read. This commit uses switch statement to evaluate the result. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: remove useless generation checkTakashi Sakamoto2025-09-181-8/+6
| | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use struct_size and flex_array_size in ioctl_add_descriptorThorsten Blum2025-09-171-2/+3
| | | | | | | | | | | | | | | | | | | | | | | | | | Use struct_size() to determine the memory needed for a new 'struct descriptor_resource' and flex_array_size() to calculate the number of bytes to copy from userspace. This removes the hardcoded size (4 bytes) for the 'u32 data[]' entries. No functional changes intended. Signed-off-by: Thorsten Blum <[email protected]> Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: shrink critical section of fw_card spinlock in bm_workTakashi Sakamoto2025-09-171-22/+8
| | | | | | | | | | | | | | | | | | | | Now fw_core_handle_bus_reset() and bm_work() are serialized. Some members of fw_card are free to access in bm_work() This commit shrinks critical section of fw_card spinlock in bm_work() Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: disable bus management work temporarily during updating topologyTakashi Sakamoto2025-09-171-0/+8
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: schedule bm_work item outside of spin lockTakashi Sakamoto2025-09-171-1/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: annotate fw_destroy_nodes with must-hold-lockTakashi Sakamoto2025-09-151-1/+3
| | | | | | | | | | | | | | | | | | | | | | The function, fw_destroy_nodes(), is used widely within firewire-core module. It has a prerequisite condition that struct fw_card.lock must be hold in advance. This commit adds annotation for it. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use spin lock specific to timer for split transactionTakashi Sakamoto2025-09-152-25/+48
| | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use spin lock specific to transactionTakashi Sakamoto2025-09-152-27/+44
| | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use spin lock specific to topology mapTakashi Sakamoto2025-09-152-9/+19
| | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: maintain phy packet receivers locally in cdev layerTakashi Sakamoto2025-09-152-8/+20
| | | | | | | | | | | | | | | | | | | | The list of receivers for phy packet is used only by cdev layer, while it is maintained as a member of fw_card structure. This commit maintains the list locally in cdev layer. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use scoped_guard() to manage critical section to update topologyTakashi Sakamoto2025-09-151-36/+30
| | | | | | | | | | | | | | | | | | | | | | At present, guard() macro is used for the critical section to update topology. It is inconvenient to add the other critical sections into the function. This commit uses scoped_guard() macro instead. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use helper macro to compare against current jiffiesTakashi Sakamoto2025-09-153-5/+4
| | | | | | | | | | | | | | | | The pattern of calling either time_before64() or time_after64() with get_jiffies_64() can be replaced with the corresponding helper macros. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use helper macros instead of direct access to HZTakashi Sakamoto2025-09-155-17/+16
| | | | | | | | | | | | | | | | There are some macros available to convert usecs, msecs, and secs into jiffies count. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use macro expression for not-registered state of BUS_MANAGER_IDTakashi Sakamoto2025-09-132-4/+11
| | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use macro expression for gap count mismatchTakashi Sakamoto2025-09-133-2/+5
| | | | | | | | | | | | | | | | | | | | | | The gap_count field is assigned to zero when mismatch is detected. In such case, the macro expression is preferable since it is easy to understand the situation. This commit applies the idea. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: remove useless lockdep_assert_held()Takashi Sakamoto2025-09-111-2/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | The bm_work work item should be scheduled after holding fw_card reference counting. At a commit 25feb1a96e21 ("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: 25feb1a96e21 ("firewire: core: use cleanup function in bm_work") Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: code refactoring whether root node is cycle master capableTakashi Sakamoto2025-09-081-5/+3
| | | | | | | | | | | | | | | | | | | | | | The check of cycle master capability in root node is currently just in a condition branch. In this case, the required variable should be within the branch. This commit is just for the purpose. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: code refactoring to investigate root node for bus managerTakashi Sakamoto2025-09-081-29/+27
| | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: code refactoring to detect both IEEE 1394:1995 IRM and Canon ↵Takashi Sakamoto2025-09-081-11/+11
| | | | | | | | | | | | | | | | | | | | | | | | | | MV5i The detection of IEEE 1394:1995 and Canon MV5i is just required within some of the condition branches. In this case, these check can be capsulated within these branches. This commit refactors the checks. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: refer fw_card member to initiate bus reset under acquiring lockTakashi Sakamoto2025-09-081-26/+26
| | | | | | | | | | | | | | | | | | | | | | The gap_count member of fw_card structure is referred when initiate bus reset. This reference is done out of acquiring lock. This is not good. This commit takes the reference within the acquiring lock, with additional code refactoring. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: code refactoring to evaluate transaction result to ↵Takashi Sakamoto2025-09-081-14/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | CSR_BUS_MANAGER_ID 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 27310d561622 ("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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: localize transaction data and rcode per condition branchTakashi Sakamoto2025-09-081-10/+11
| | | | | | | | | | | | | | | | | | | | | | The function local variable, transaction_data, in bm_work function is conditionally used. In the case, the branch-level variable is sometimes useful. This commit uses this idea. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: use cleanup function in bm_workTakashi Sakamoto2025-09-081-17/+16
| | | | | | | | | | | | | | | | | | | | | | In "bm_work" function, the references to fw_card and fw_node are released at last. This is achieved by using goto statements. For this case, the kernel cleanup framework is available. This commit uses the framework to remove these statements. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: add helper functions to access to fw_device data in fw_node ↵Takashi Sakamoto2025-09-083-13/+23
| | | | | | | | | | | | | | | | | | | | | | | | | | structure 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: use return value from fw_node_get()Takashi Sakamoto2025-09-081-2/+1
| | | | | | | | | | | | | | | | | | | | The programming pattern, referring after increasing reference count, is supported by fw_node_get(). This commit simplify the programming pattern. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: core: utilize cleanup function to release workqueue in error pathTakashi Sakamoto2025-09-081-18/+22
| | | | | | | | | | | | | | | | | | | | | | The helper macro, retain_and_null_ptr(), introduced by a commit 092d00ead733 ("cleanup: Provide retain_and_null_ptr()") in v6.16 kernel, is useful in the error path to release the part of structure member. This commit uses the relatively new function. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: use kcalloc() variant for array allocationTakashi Sakamoto2025-09-081-5/+2
| | | | | | | | | | | | | | | | | | | | | | When allocating the list of isochronous context structure, a kzalloc() variant of managed device API is used. In this case, a kcalloc() variant is available. This commit replaces these lines with devm_kcalloc(). Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: remove module-local workqueueTakashi Sakamoto2025-08-231-11/+0
| | | | | | | | | | | | | | | | | | | | Now module-local workqueue has been replaced by a threaded IRQ handler. This commit removes the workqueue and the associated work item accordingly. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: use threaded IRQ handler to handle SelfIDComplete eventTakashi Sakamoto2025-08-231-18/+22
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: move self_id_complete tracepoint after validating registerTakashi Sakamoto2025-08-231-8/+4
| | | | | | | | | | | | | | | | | | The value of OHCI1394_SelfIDCount register includes an error-indicating bit. It is safer to place the tracepoint probe after validating the register value. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: remove obsolete module-level debug parameterTakashi Sakamoto2025-08-221-10/+0
| | | | | | | | | | | | | | | | | | | | | | The module-level debug parameter was added in v2.6.26 by a commit ad3c0fe8b8d16 ("firewire: debug interrupt events"). Its functionality has long been superseded by tracepoints. This commit removes the module parameter, bye. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: remove obsolete debug logging for AT/AR resultsTakashi Sakamoto2025-08-221-113/+19
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: remove obsolete debug logging for selfID sequenceTakashi Sakamoto2025-08-221-72/+0
| | | | | | | | | | | | | | | | | | | | | | | | | | | | A commit 677ceae19073 ("firewire: core: add tracepoints event for self_id_sequence") added the "firewire:self_id_sequence" event in v6.11. A commit 526e21a2aa6f ("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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
| * firewire: ohci: remove obsolete debug logging for IRQ eventsTakashi Sakamoto2025-08-221-33/+1
| | | | | | | | | | | | | | | | | | | | | | A commit 0d8914165dd1 ("firewire: ohci: add tracepoints event for hardIRQ event") added "firewire_ohci:irqs" event in v6.11, which can provide equivalent information to the existing debug logging. This commit removes the logging. Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
* | firewire: core: fix overlooked update of subsystem ABI versionTakashi Sakamoto2025-09-201-1/+1
|/ | | | | | | | | | | | 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: 6add87e9764d ("firewire: cdev: add new version of ABI to notify time stamp at request/response subaction of transaction#") Link: https://lore.kernel.org/r/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
* firewire: core: reallocate buffer for FCP address handlers when more than 4 ↵Takashi Sakamoto2025-08-031-4/+32
| | | | | | | | | | | | | | | | | are registered 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/[email protected] Signed-off-by: Takashi Sakamoto <[email protected]>
generated by cgit (git 2.51.2) at 2025-12-04 01:15:46 +0000