[PATCH] bus: mhi: host: Change the trace string for the userspace tools mapping
User space tools can't map strings if we use directly, as the string address is internal to kernel. So add trace point strings for the user space tools to map strings properly. Signed-off-by: Krishna chaitanya chundru --- drivers/bus/mhi/host/main.c | 4 ++-- drivers/bus/mhi/host/trace.h | 2 ++ 2 files changed, 4 insertions(+), 2 deletions(-) diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index 2d38f6005da6..15d657af9b5b 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -1340,7 +1340,7 @@ static int mhi_update_channel_state(struct mhi_controller *mhi_cntrl, enum mhi_cmd_type cmd = MHI_CMD_NOP; int ret; - trace_mhi_channel_command_start(mhi_cntrl, mhi_chan, to_state, "Updating"); + trace_mhi_channel_command_start(mhi_cntrl, mhi_chan, to_state, TPS("Updating")); switch (to_state) { case MHI_CH_STATE_TYPE_RESET: write_lock_irq(_chan->lock); @@ -1407,7 +1407,7 @@ static int mhi_update_channel_state(struct mhi_controller *mhi_cntrl, write_unlock_irq(_chan->lock); } - trace_mhi_channel_command_end(mhi_cntrl, mhi_chan, to_state, "Updated"); + trace_mhi_channel_command_end(mhi_cntrl, mhi_chan, to_state, TPS("Updated")); exit_channel_update: mhi_cntrl->runtime_put(mhi_cntrl); mhi_device_put(mhi_cntrl->mhi_dev); diff --git a/drivers/bus/mhi/host/trace.h b/drivers/bus/mhi/host/trace.h index d12a98d44272..368515dcb22d 100644 --- a/drivers/bus/mhi/host/trace.h +++ b/drivers/bus/mhi/host/trace.h @@ -84,6 +84,8 @@ DEV_ST_TRANSITION_LIST #define dev_st_trans(a, b) { DEV_ST_TRANSITION_##a, b }, #define dev_st_trans_end(a, b) { DEV_ST_TRANSITION_##a, b } +#define TPS(x) tracepoint_string(x) + TRACE_EVENT(mhi_gen_tre, TP_PROTO(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan, --- base-commit: ceeb64f41fe6a1eb9fc56d583983a81f8f3dd058 change-id: 20240218-ftrace_string-7677762aa63c Best regards, -- Krishna chaitanya chundru
Re: [PATCH v11] bus: mhi: host: Add tracing support
On 2/6/2024 11:56 AM, Manivannan Sadhasivam wrote: On Tue, Feb 06, 2024 at 10:02:05AM +0530, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Change the implementation of the arrays which has enum to strings mapping to make it consistent in both trace header file and other files. Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru There are a lot of checkpatch errors. Please fix them and resubmit. - Mani Hi Mani, The errors which is pointing in the checkpatch are false positive, those errors are being shown from v2 onwards and kernel test boot is also not throwing any errors for it. I checked with internal team they said these errors are false positive. Thanks & Regards, Krishna Chaitanya. Reviewed-by: Manivannan Sadhasivam Reviewed-by: "Steven Rostedt (Google)" --- Changes in v11: - Rebased with mhi next. - Link to v10: https://lore.kernel.org/r/20240131-ftrace_support-v10-1-4349306b8...@quicinc.com Changes in v10: - Modified command_start and command_end traces to take string as input to mention correct - string as suggested by mani - As sugggested by mani modified the print format from lower format to upper case format. - Link to v9: https://lore.kernel.org/r/20240105-ftrace_support-v9-1-a2dca64cc...@quicinc.com Changes in v9: - Change the implementations of some array so that the strings to enum mapping - is same in both trace header and other files as suggested by steve. - Link to v8: https://lore.kernel.org/r/20231207-ftrace_support-v8-1-7f62d4558...@quicinc.com Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/common.h| 38 +++--- drivers/bus/mhi/host/init.c | 64 + drivers/bus/mhi/host/internal.h | 41 ++ drivers/bus/mhi/host/main.c | 19 ++- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h| 280 6 files changed, 384 insertions(+), 65 deletions(-) diff --git a/drivers/bus/mhi/common.h b/drivers/bus/mhi/common.h index f794b9c8049e..dda340aaed95 100644 --- a/drivers/bus/mhi/common.h +++ b/drivers/bus/mhi/common.h @@ -297,30 +297,30 @@ struct mhi_ring_element { __le32 dword[2]; }; +#define MHI_STATE_LIST\ + mhi_state(RESET,"RESET") \ + mhi_state(READY,"READY") \ + mhi_state(M0, "M0") \ + mhi_state(M1, "M1") \ + mhi_state(M2, "M2") \ + mhi_state(M3, "M3") \ + mhi_state(M3_FAST, "M3_FAST")\ + mhi_state(BHI,
[PATCH v11] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Change the implementation of the arrays which has enum to strings mapping to make it consistent in both trace header file and other files. Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru Reviewed-by: Manivannan Sadhasivam Reviewed-by: "Steven Rostedt (Google)" --- Changes in v11: - Rebased with mhi next. - Link to v10: https://lore.kernel.org/r/20240131-ftrace_support-v10-1-4349306b8...@quicinc.com Changes in v10: - Modified command_start and command_end traces to take string as input to mention correct - string as suggested by mani - As sugggested by mani modified the print format from lower format to upper case format. - Link to v9: https://lore.kernel.org/r/20240105-ftrace_support-v9-1-a2dca64cc...@quicinc.com Changes in v9: - Change the implementations of some array so that the strings to enum mapping - is same in both trace header and other files as suggested by steve. - Link to v8: https://lore.kernel.org/r/20231207-ftrace_support-v8-1-7f62d4558...@quicinc.com Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/common.h| 38 +++--- drivers/bus/mhi/host/init.c | 64 + drivers/bus/mhi/host/internal.h | 41 ++ drivers/bus/mhi/host/main.c | 19 ++- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h| 280 6 files changed, 384 insertions(+), 65 deletions(-) diff --git a/drivers/bus/mhi/common.h b/drivers/bus/mhi/common.h index f794b9c8049e..dda340aaed95 100644 --- a/drivers/bus/mhi/common.h +++ b/drivers/bus/mhi/common.h @@ -297,30 +297,30 @@ struct mhi_ring_element { __le32 dword[2]; }; +#define MHI_STATE_LIST \ + mhi_state(RESET,"RESET")\ + mhi_state(READY,"READY")\ + mhi_state(M0, "M0") \ + mhi_state(M1, "M1") \ + mhi_state(M2, "M2") \ + mhi_state(M3, "M3") \ + mhi_state(M3_FAST, "M3_FAST") \ + mhi_state(BHI, "BHI") \ + mhi_state_end(SYS_ERR, "SYS ERROR") + +#undef mhi_state +#undef mhi_state_end + +#define mhi_state(a, b)case MHI_STATE_##a: return b; +#define mhi_state_end(a, b)case MHI_STATE_##a: return b; + static inline const char *mhi_state_str(enum mhi_state state) { switch (state) { - case MHI_STATE_RESET: - return "RESET"; - case MHI_STATE_READY: - return &quo
[PATCH v10] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Change the implementation of the arrays which has enum to strings mapping to make it consistent in both trace header file and other files. Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru Reviewed-by: "Steven Rostedt (Google)" --- Changes in v10: - Modified command_start and command_end traces to take string as input to mention correct - string as suggested by mani - As sugggested by mani modified the print format from lower format to upper case format. - Link to v9: https://lore.kernel.org/r/20240105-ftrace_support-v9-1-a2dca64cc...@quicinc.com Changes in v9: - Change the implementations of some array so that the strings to enum mapping - is same in both trace header and other files as suggested by steve. - Link to v8: https://lore.kernel.org/r/20231207-ftrace_support-v8-1-7f62d4558...@quicinc.com Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/common.h| 38 +++--- drivers/bus/mhi/host/init.c | 63 + drivers/bus/mhi/host/internal.h | 40 ++ drivers/bus/mhi/host/main.c | 19 ++- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h| 280 6 files changed, 383 insertions(+), 64 deletions(-) diff --git a/drivers/bus/mhi/common.h b/drivers/bus/mhi/common.h index f794b9c8049e..dda340aaed95 100644 --- a/drivers/bus/mhi/common.h +++ b/drivers/bus/mhi/common.h @@ -297,30 +297,30 @@ struct mhi_ring_element { __le32 dword[2]; }; +#define MHI_STATE_LIST \ + mhi_state(RESET,"RESET")\ + mhi_state(READY,"READY")\ + mhi_state(M0, "M0") \ + mhi_state(M1, "M1") \ + mhi_state(M2, "M2") \ + mhi_state(M3, "M3") \ + mhi_state(M3_FAST, "M3_FAST") \ + mhi_state(BHI, "BHI") \ + mhi_state_end(SYS_ERR, "SYS ERROR") + +#undef mhi_state +#undef mhi_state_end + +#define mhi_state(a, b)case MHI_STATE_##a: return b; +#define mhi_state_end(a, b)case MHI_STATE_##a: return b; + static inline const char *mhi_state_str(enum mhi_state state) { switch (state) { - case MHI_STATE_RESET: - return "RESET"; - case MHI_STATE_READY: - return "READY"; - case MHI_STATE_M0: - return "M0"; - case MHI_STATE_M1: - return "M1"; - case MHI_STATE_M2: -
Re: [PATCH v9] bus: mhi: host: Add tracing support
On 1/30/2024 11:56 PM, Manivannan Sadhasivam wrote: On Tue, Jan 30, 2024 at 09:22:52AM -0500, Steven Rostedt wrote: On Tue, 30 Jan 2024 13:41:52 +0530 Manivannan Sadhasivam wrote: So same trace will get printed for both mhi_channel_command_start() and mhi_channel_command_end()? The trace output will also include the tracepoint name. That is, it will have the same content but will be preceded with: mhi_channel_command_start: ... mhi_channel_command_end: ... Yes, but the message will be the same: mhi_channel_command_start: chan%d: Updating state to: mhi_channel_command_end: chan%d: Updating state to: Either only one of the trace should be present or the second one should print, "mhi_channel_command_end: chan%d: Updated state to:" - Mani I will try to pass a string to updated for mhi_channel_command_end & updating for mhi_channel_command_start in my next patch. - Krishna Chaitanya.
Re: [PATCH v9] bus: mhi: host: Add tracing support
On 1/30/2024 1:41 PM, Manivannan Sadhasivam wrote: On Fri, Jan 05, 2024 at 05:53:03PM +0530, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Change the implementation of the arrays which has enum to strings mapping to make it consistent in both trace header file and other files. Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru Few nitpicks below. I will make suggested changes in next patch. - Krishna Chaitanya. Reviewed-by: "Steven Rostedt (Google)" --- Changes in v9: - Change the implementations of some array so that the strings to enum mapping - is same in both trace header and other files as suggested by steve. - Link to v8: https://lore.kernel.org/r/20231207-ftrace_support-v8-1-7f62d4558...@quicinc.com Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/common.h| 38 +++--- drivers/bus/mhi/host/init.c | 63 + drivers/bus/mhi/host/internal.h | 40 ++ drivers/bus/mhi/host/main.c | 19 ++- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h| 275 6 files changed, 378 insertions(+), 64 deletions(-) [...] +TRACE_EVENT(mhi_gen_tre, + + TP_PROTO(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan, +struct mhi_ring_element *mhi_tre), + + TP_ARGS(mhi_cntrl, mhi_chan, mhi_tre), + + TP_STRUCT__entry( + __string(name, mhi_cntrl->mhi_dev->name) + __field(int, ch_num) + __field(void *, wp) + __field(__le64, tre_ptr) + __field(__le32, dword0) + __field(__le32, dword1) + ), + + TP_fast_assign( + __assign_str(name, mhi_cntrl->mhi_dev->name); + __entry->ch_num = mhi_chan->chan; + __entry->wp = mhi_tre; + __entry->tre_ptr = mhi_tre->ptr; + __entry->dword0 = mhi_tre->dword[0]; + __entry->dword1 = mhi_tre->dword[1]; + ), + + TP_printk("%s: Chan: %d Tre: 0x%p Tre buf: 0x%llx dword0: 0x%08x dword1: 0x%08x\n", Use caps for printing the acronyms everywhere. Like TRE, DWORD etc... + __get_str(name), __entry->ch_num, __entry->wp, __entry->tre_ptr, + __entry->dword0, __entry->dword1) +); + +TRACE_EVENT(mhi_intvec_states, + + TP_PROTO(struct mhi_controller *mhi_cntrl, int dev_ee, int dev_state), + + TP_ARGS(mhi_cntrl, dev_ee, dev_state), + + TP_STRUCT__entry( + __string(name, mhi_cntrl->mhi_dev->name) + __field(in
Re: [PATCH v9] bus: mhi: host: Add tracing support
On 1/8/2024 6:52 PM, Krishna Chaitanya Chundru wrote: Hi Steven, Even though I added your reviewed-by tag, I incorporated changes mentioned in the previous patch. Can you please review it once. Thanks & Regards, Krishna Chaitanya. Hi Steven, Can you please review it once. Thanks & Regards, Krishna Chaitanya. On 1/5/2024 5:53 PM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Change the implementation of the arrays which has enum to strings mapping to make it consistent in both trace header file and other files. Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru Reviewed-by: "Steven Rostedt (Google)" --- Changes in v9: - Change the implementations of some array so that the strings to enum mapping - is same in both trace header and other files as suggested by steve. - Link to v8: https://lore.kernel.org/r/20231207-ftrace_support-v8-1-7f62d4558...@quicinc.com Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/common.h | 38 +++--- drivers/bus/mhi/host/init.c | 63 + drivers/bus/mhi/host/internal.h | 40 ++ drivers/bus/mhi/host/main.c | 19 ++- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h | 275 6 files changed, 378 insertions(+), 64 deletions(-) diff --git a/drivers/bus/mhi/common.h b/drivers/bus/mhi/common.h index f794b9c8049e..dda340aaed95 100644 --- a/drivers/bus/mhi/common.h +++ b/drivers/bus/mhi/common.h @@ -297,30 +297,30 @@ struct mhi_ring_element { __le32 dword[2]; }; +#define MHI_STATE_LIST \ + mhi_state(RESET, "RESET") \ + mhi_state(READY, "READY") \ + mhi_state(M0, "M0") \ + mhi_state(M1, "M1") \ + mhi_state(M2, "M2") \ + mhi_state(M3, "M3") \ + mhi_state(M3_FAST, "M3_FAST") \ + mhi_state(BHI, "BHI") \ + mhi_state_end(SYS_ERR, "SYS ERROR") + +#undef mhi_state +#undef mhi_state_end + +#define mhi_state(a, b) case MHI_STATE_##a: return b; +#define mhi_state_end(a, b) case MHI_STATE_##a: return b; + static inline const char *mhi_state_str(enum mhi_state state) { switch (state) { - case MHI_STATE_RESET: - return "RESET"; - case MHI_STATE_READY: - return "READY"; - case MHI_STATE_M0: - return "M0"; - case MHI_STATE_M1: - return "M1"; - case MHI_STATE_M2: - return "M2"
Re: [PATCH v9] bus: mhi: host: Add tracing support
Hi Steven, Even though I added your reviewed-by tag, I incorporated changes mentioned in the previous patch. Can you please review it once. Thanks & Regards, Krishna Chaitanya. On 1/5/2024 5:53 PM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Change the implementation of the arrays which has enum to strings mapping to make it consistent in both trace header file and other files. Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru Reviewed-by: "Steven Rostedt (Google)" --- Changes in v9: - Change the implementations of some array so that the strings to enum mapping - is same in both trace header and other files as suggested by steve. - Link to v8: https://lore.kernel.org/r/20231207-ftrace_support-v8-1-7f62d4558...@quicinc.com Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/common.h| 38 +++--- drivers/bus/mhi/host/init.c | 63 + drivers/bus/mhi/host/internal.h | 40 ++ drivers/bus/mhi/host/main.c | 19 ++- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h| 275 6 files changed, 378 insertions(+), 64 deletions(-) diff --git a/drivers/bus/mhi/common.h b/drivers/bus/mhi/common.h index f794b9c8049e..dda340aaed95 100644 --- a/drivers/bus/mhi/common.h +++ b/drivers/bus/mhi/common.h @@ -297,30 +297,30 @@ struct mhi_ring_element { __le32 dword[2]; }; +#define MHI_STATE_LIST\ + mhi_state(RESET,"RESET") \ + mhi_state(READY,"READY") \ + mhi_state(M0, "M0") \ + mhi_state(M1, "M1") \ + mhi_state(M2, "M2") \ + mhi_state(M3, "M3") \ + mhi_state(M3_FAST, "M3_FAST")\ + mhi_state(BHI, "BHI")\ + mhi_state_end(SYS_ERR, "SYS ERROR") + +#undef mhi_state +#undef mhi_state_end + +#define mhi_state(a, b)case MHI_STATE_##a: return b; +#define mhi_state_end(a, b)case MHI_STATE_##a: return b; + static inline const char *mhi_state_str(enum mhi_state state) { switch (state) { - case MHI_STATE_RESET: - return "RESET"; - case MHI_STATE_READY: - return "READY"; - case MHI_STATE_M0: - return "M0"; - case MHI_STATE_M1: - return "M1"; - case MHI_STATE_M2: - return "M2"; - case MHI_STATE_M3: - return &q
[PATCH v9] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Change the implementation of the arrays which has enum to strings mapping to make it consistent in both trace header file and other files. Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru Reviewed-by: "Steven Rostedt (Google)" --- Changes in v9: - Change the implementations of some array so that the strings to enum mapping - is same in both trace header and other files as suggested by steve. - Link to v8: https://lore.kernel.org/r/20231207-ftrace_support-v8-1-7f62d4558...@quicinc.com Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/common.h| 38 +++--- drivers/bus/mhi/host/init.c | 63 + drivers/bus/mhi/host/internal.h | 40 ++ drivers/bus/mhi/host/main.c | 19 ++- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h| 275 6 files changed, 378 insertions(+), 64 deletions(-) diff --git a/drivers/bus/mhi/common.h b/drivers/bus/mhi/common.h index f794b9c8049e..dda340aaed95 100644 --- a/drivers/bus/mhi/common.h +++ b/drivers/bus/mhi/common.h @@ -297,30 +297,30 @@ struct mhi_ring_element { __le32 dword[2]; }; +#define MHI_STATE_LIST \ + mhi_state(RESET,"RESET")\ + mhi_state(READY,"READY")\ + mhi_state(M0, "M0") \ + mhi_state(M1, "M1") \ + mhi_state(M2, "M2") \ + mhi_state(M3, "M3") \ + mhi_state(M3_FAST, "M3_FAST") \ + mhi_state(BHI, "BHI") \ + mhi_state_end(SYS_ERR, "SYS ERROR") + +#undef mhi_state +#undef mhi_state_end + +#define mhi_state(a, b)case MHI_STATE_##a: return b; +#define mhi_state_end(a, b)case MHI_STATE_##a: return b; + static inline const char *mhi_state_str(enum mhi_state state) { switch (state) { - case MHI_STATE_RESET: - return "RESET"; - case MHI_STATE_READY: - return "READY"; - case MHI_STATE_M0: - return "M0"; - case MHI_STATE_M1: - return "M1"; - case MHI_STATE_M2: - return "M2"; - case MHI_STATE_M3: - return "M3"; - case MHI_STATE_M3_FAST: - return "M3 FAST"; - case MHI_STATE_BHI: - return "BHI"; - case MHI_STATE_SYS_ERR: - return "SYS ERROR"; +
Re: [PATCH v8] bus: mhi: host: Add tracing support
On 1/5/2024 10:47 AM, Baochen Qiang wrote: On 1/4/2024 12:47 PM, Krishna Chaitanya Chundru wrote: Hi Steven, Can you please review this. Thanks & Regards, Krishna Chaitanya. On 12/7/2023 10:00 AM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Is there a reason why debug messages have to be removed? From the view of MHI controller, we often need MHI logs to debug, so is it possbile to preserve those logs? The trace is being replaced by the debug message and it would be preferred to have one mechanism or the other, not both. you will still see these logs in traces. - Krishna Chaitanya. Signed-off-by: Krishna chaitanya chundru --- Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/main.c | 19 ++-- drivers/bus/mhi/host/pm.c | 7 +- drivers/bus/mhi/host/trace.h | 205 +++ 4 files changed, 221 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..189f4786403e 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -15,6 +15,7 @@ #include #include #include "internal.h" +#include "trace.h" int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl, void __iomem *base, u32 offset, u32 *out) @@ -491,11 +492,8 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,8 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl, local_rp); + switc
Re: [PATCH v8] bus: mhi: host: Add tracing support
On 1/4/2024 9:31 PM, Steven Rostedt wrote: On Thu, 7 Dec 2023 10:00:47 +0530 Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com So this looks good from a tracing POV. Reviewed-by: Steven Rostedt (Google) But I do have some more comments that could be done by new patches. +TRACE_EVENT(mhi_intvec_states, + + TP_PROTO(struct mhi_controller *mhi_cntrl, int dev_ee, int dev_state), + + TP_ARGS(mhi_cntrl, dev_ee, dev_state), + + TP_STRUCT__entry( + __string(name, mhi_cntrl->mhi_dev->name) + __field(int, local_ee) + __field(int, state) + __field(int, dev_ee) + __field(int, dev_state) + ), + + TP_fast_assign( + __assign_str(name, mhi_cntrl->mhi_dev->name); + __entry->local_ee = mhi_cntrl->ee; + __entry->state = mhi_cntrl->dev_state; + __entry->dev_ee = dev_ee; + __entry->dev_state = dev_state; + ), + + TP_printk("%s: local ee: %s state: %s device ee: %s state: %s\n", + __get_str(name), + TO_MHI_EXEC_STR(__entry->local_ee), + mhi_state_str(__entry->state), + TO_MHI_EXEC_STR(__entry->dev_ee), + mhi_state_str(__entry->dev_state)) So the above may have issues with user space parsing. For one, that mhi_state_str() is: static inline const char *mhi_state_str(enum mhi_state state) { switch (state) { case MHI_STATE_RESET: return "RESET"; case MHI_STATE_READY: return "READY"; case MHI_STATE_M0: return "M0"; case MHI_STATE_M1: return "M1"; case MHI_STATE_M2: return "M2"; case MHI_STATE_M3: return "M3"; case MHI_STATE_M3_FAST: return "M3 FAST"; case MHI_STATE_BHI: return "BHI"; case MHI_STATE_SYS_ERR: return "SYS ERROR"; default: return "Unknown state"; } }; Which if this could be changed into: #define MHI_STATE_LIST \ EM(RESET,"RESET") \ EM(READY,"READY") \ EM(M0, "M0")\ EM(M1, "M1")\ EM(M2, "M2")\ EM(M3, "M3")\ EM(M3_FAST, "M3_FAST") \ EM(BHI, "BHI") \ EMe(SYS_ERR, "SYS ERROR") #undef EM #undef EMe #define EM(a, b) case MHI_STATE_##a: return b; #define EMe(a, b) case MHI_STATE_##a: return b; static inline const char *mhi_state_str(enum mhi_state state) { switch (state) { MHI_STATE_LIST default: return "Unknown state"; } Then you could use that macro in the trace header: #undef EM #undef EMe #define EM(a, b)TRACE_DEFINE_ENUM(MHI_STATE_##a); #define EMe(a, b) TRACE_DEFINE_ENUM(MHI_STATE_##a); MHI_STATE_LIST And in the print fmts: #undef EM #undef EMe #define EM(a, b) { MHI_STATE_##a, b }, #define EMe(a, b) { MHI_STATE_##a, b } TP_printk("%s: local ee: %s state: %s device ee: %s state: %s\n", __get_str(name), TO_MHI_EXEC_STR(__entry->local_ee), __print_symbolic(__entry->state), MHI_STATE_LIST), TO_MHI_EXEC_STR(__entry->dev_ee), __print_symbolic(__entry->dev_state, MHI_STATE_LIST)) And that will be exported to user space in the /sys/kernel/tracing/events/*/*/format file, as: __print_symbolic(REC->state, { { MHI_STATE_RESET, "RESET"}, { MHI_STATE_READY, "READY"}, { MHI_STATE_M0, "M0"}, { MHI_STATE_M1, "M1"}, { MHI_STATE_M2, "M2"}, { MHI_STATE_M3, "M3"}, { MHI_STATE_M3_FAST, "M3 FAST&qu
Re: [PATCH v8] bus: mhi: host: Add tracing support
Hi Steven, Can you please review this. Thanks & Regards, Krishna Chaitanya. On 12/7/2023 10:00 AM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/main.c | 19 ++-- drivers/bus/mhi/host/pm.c| 7 +- drivers/bus/mhi/host/trace.h | 205 +++ 4 files changed, 221 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..189f4786403e 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -15,6 +15,7 @@ #include #include #include "internal.h" +#include "trace.h" int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl, void __iomem *base, u32 offset, u32 *out) @@ -491,11 +492,8 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,8 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl, local_rp); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +997,8 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl, local_rp); + chan = MHI_TRE_GET_EV_CH
Re: [PATCH v8] bus: mhi: host: Add tracing support
Hi Steven, Can you review it once. Thanks & Regards, Krishna Chaitanya. On 12/7/2023 10:00 AM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/main.c | 19 ++-- drivers/bus/mhi/host/pm.c| 7 +- drivers/bus/mhi/host/trace.h | 205 +++ 4 files changed, 221 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..189f4786403e 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -15,6 +15,7 @@ #include #include #include "internal.h" +#include "trace.h" int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl, void __iomem *base, u32 offset, u32 *out) @@ -491,11 +492,8 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,8 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl, local_rp); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +997,8 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl, local_rp); + chan = MHI_TRE_GET_EV_CHID(local_rp)
[PATCH v8] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v8: - Pass the structure and derefernce the variables in TP_fast_assign as suggested by steve - Link to v7: https://lore.kernel.org/r/20231206-ftrace_support-v7-1-aca49a042...@quicinc.com Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/main.c | 19 ++-- drivers/bus/mhi/host/pm.c| 7 +- drivers/bus/mhi/host/trace.h | 205 +++ 4 files changed, 221 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..189f4786403e 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -15,6 +15,7 @@ #include #include #include "internal.h" +#include "trace.h" int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl, void __iomem *base, u32 offset, u32 *out) @@ -491,11 +492,8 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,8 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl, local_rp); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +997,8 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl, local_rp); + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1237,7 @@ int mhi_gen_tre(struct mhi_controller *mhi_
Re: [PATCH v7] bus: mhi: host: Add tracing support
On 12/6/2023 9:25 PM, Steven Rostedt wrote: On Wed, 6 Dec 2023 21:12:57 +0530 Krishna chaitanya chundru wrote: diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..507cf3351a97 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -15,6 +15,7 @@ #include #include #include "internal.h" +#include "trace.h" int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl, void __iomem *base, u32 offset, u32 *out) @@ -491,11 +492,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); I'm not sure if this was discussed before, but why not just pass in the mhi_cntrl pointer and do the assigning in the TRACE_EVENT() TP_fast_assign() portion? It will save on setting up the parameters. if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +831,10 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl->mhi_dev->name, local_rp, +local_rp->ptr, local_rp->dword[0], +local_rp->dword[1]); And here.. + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +1000,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl->mhi_dev->name, local_rp, local_rp->ptr, +local_rp->dword[0], local_rp->dword[1]); and here.. for the local_rp. Especially since you are just repeating what you send into the two events. You can consolidate it down to just adding that in the TP_fast_assign() of the shared DECLARE_EVENT_CLASS(). + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1241,8 @@ int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan, mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(info->len); mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, eot, eob, chain); + trace_mhi_gen_tre(mhi_cntrl->mhi_dev->name, mhi_chan->chan, mhi_tre, + mhi_tre->ptr, mhi_tre->dword[0], mhi_tre->dword[1]); Same here. You can pass in the mhi_tre pointer. /* increment WP */ mhi_add_ring_element(mhi_cntrl, tre_ring); mhi_add_ring_element(mhi_cntrl, buf_ring); @@ -1327,9 +1335,7 @@ static int mhi_update_channel_state(struct mhi_controller *mhi_cntrl, enum mhi_cmd_type cmd = MHI_CMD_NOP; int ret; - dev_dbg(dev, "%d: Updating channel state to: %s\n", mhi_chan->chan, - TO_CH_STATE_TYPE_STR(to_state)); - + trace_mhi_channel_command_start(mhi_cntrl->mhi_dev->name, mhi_chan->chan, to_state); And here.. switch (to_state) { case MHI_CH_STATE_TYPE_RESET: write_lock_irq(_chan->lock); @@ -1396,9 +1402,7 @@ static int mhi_update_channel_state(struct mhi_controller *mhi_cntrl, write_unlock_irq(_chan->lock); } - dev_dbg(dev, "%d: Channel state change to %s successful\n", - mhi_chan->chan, TO_CH_STATE_TYPE_STR(to_state)); - + trace_mhi_channel_command_end(mhi_cntrl->mhi_dev->name, mhi_chan->chan, to_state); and here.. Where you can also update the DECLARE_EVENT_CLASS() to directly access the mhi_cntrl and mhi_chan pointers. Sometimes it's better to do the dereference from inside the TP_fast_assign. That way, things like eprobes and bpf tracing can also have access to the full structure if needed. -- Steve Hi Steve, I will make changes as suggested in the next patch series. - Krishna Chaitanya.
[PATCH v7] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v7: - change log format as pointed by mani. - Link to v6: https://lore.kernel.org/r/20231204-ftrace_support-v6-1-9b206546d...@quicinc.com Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/main.c | 24 ++--- drivers/bus/mhi/host/pm.c| 7 +- drivers/bus/mhi/host/trace.h | 208 +++ 4 files changed, 229 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..507cf3351a97 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -15,6 +15,7 @@ #include #include #include "internal.h" +#include "trace.h" int __must_check mhi_read_reg(struct mhi_controller *mhi_cntrl, void __iomem *base, u32 offset, u32 *out) @@ -491,11 +492,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +831,10 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl->mhi_dev->name, local_rp, +local_rp->ptr, local_rp->dword[0], +local_rp->dword[1]); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +1000,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl->mhi_dev->name, local_rp, local_rp->ptr, +local_rp->dword[0], local_rp->dword[1]);
[PATCH v6] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v6: - use 'rp' directly as suggested by jeffrey. - Link to v5: https://lore.kernel.org/r/20231127-ftrace_support-v5-1-eb67daead...@quicinc.com Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 23 +++-- drivers/bus/mhi/host/pm.c | 6 +- drivers/bus/mhi/host/trace.h| 208 5 files changed, 228 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a02a71605907 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include "trace.h" #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..0d7e068e713a 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -491,11 +491,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,10 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl->mhi_dev->name, local_rp, +local_rp->ptr, local_rp->dword[0], +local_rp->dword[1]); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +999,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl->mhi_dev->name, local_rp, local_rp->ptr, +local_rp->dword[0], local_rp-&
Re: [PATCH v5] bus: mhi: host: Add tracing support
On 12/1/2023 10:31 PM, Jeffrey Hugo wrote: On 11/27/2023 4:09 AM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 23 +++-- drivers/bus/mhi/host/pm.c | 6 +- drivers/bus/mhi/host/trace.h | 208 5 files changed, 228 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a02a71605907 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include "trace.h" #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..0d7e068e713a 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -491,11 +491,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,10 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl->mhi_dev->name, local_rp, + local_rp->ptr, local_rp->dword[0], + local_rp->dword[1]); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +999,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl->mhi_dev->name, local_rp, local_rp->ptr, + local_rp->dword[0], local_rp->dword[1]); + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1240,8 @@ int mhi_gen_tre(struct mhi_con
[PATCH v5] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever the trace events are added, debug messages are removed. Signed-off-by: Krishna chaitanya chundru --- Changes in v5: - Use DECLARE_EVENT_CLASS for multiple events as suggested by steve. - Instead of converting to u64 to print address, use %px to print the address to avoid - warnings in some platforms. - Link to v4: https://lore.kernel.org/r/2023-ftrace_support-v4-1-c83602399...@quicinc.com Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 23 +++-- drivers/bus/mhi/host/pm.c | 6 +- drivers/bus/mhi/host/trace.h| 208 5 files changed, 228 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a02a71605907 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include "trace.h" #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..0d7e068e713a 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -491,11 +491,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,10 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_ctrl_event(mhi_cntrl->mhi_dev->name, local_rp, +local_rp->ptr, local_rp->dword[0], +local_rp->dword[1]); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +999,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_data_event(mhi_cntrl->mhi_dev->name, local_rp, local_rp->ptr, +local_rp->dword[0], local_rp->dword[1]); + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1240,8 @@ int mh
Re: [PATCH v4] bus: mhi: host: Add tracing support
On 11/12/2023 1:07 AM, Steven Rostedt wrote: On Sat, 11 Nov 2023 11:25:22 +0530 Krishna chaitanya chundru wrote: diff --git a/drivers/bus/mhi/host/trace.h b/drivers/bus/mhi/host/trace.h new file mode 100644 index ..0e99318f5d08 --- /dev/null +++ b/drivers/bus/mhi/host/trace.h + +TRACE_EVENT(mhi_update_channel_state_start, + + TP_PROTO(const char *name, int ch_num, int state), + + TP_ARGS(name, ch_num, state), + + TP_STRUCT__entry( + __string(name, name) + __field(int, ch_num) + __field(int, state) + ), + + TP_fast_assign( + __assign_str(name, name); + __entry->ch_num = ch_num; + __entry->state = state; + ), + + TP_printk("%s: ch%d: Updating state to: %s\n", + __get_str(name), __entry->ch_num, + TO_CH_STATE_TYPE_STR(__entry->state)) +); + +TRACE_EVENT(mhi_update_channel_state_end, + + TP_PROTO(const char *name, int ch_num, int state), + + TP_ARGS(name, ch_num, state), + + TP_STRUCT__entry( + __string(name, name) + __field(int, ch_num) + __field(int, state) + ), + + TP_fast_assign( + __assign_str(name, name); + __entry->ch_num = ch_num; + __entry->state = state; + ), + + TP_printk("%s: ch%d: Updated state to: %s\n", + __get_str(name), __entry->ch_num, + TO_CH_STATE_TYPE_STR(__entry->state)) +); + The above three events have the same format. You can save kilobytes of memory by converting them into a DECLARE_EVENT_CLASS() and use DEFINE_EVENT() for each event. A TRACE_EVENT() macro is really just a wrapper around DECLARE_EVENT_CLASS() and DEFINE_EVENT(). The DECLARE_EVENT_CLASS() does the bulk of the work and adds the most memory footprint. By breaking it apart for several events, it does save memory. Whenever you can use a single DECLARE_EVENT_CLASS() for multiple events, I strongly suggest doing so. Thanks, -- Steve Sure steve I will change as suggested in my next patch. - Krishna Chaitanya. +#endif +#undef TRACE_INCLUDE_PATH +#define TRACE_INCLUDE_PATH . +#undef TRACE_INCLUDE_FILE +#define TRACE_INCLUDE_FILE trace + +#include --- base-commit: 3006adf3be79cde4d14b1800b963b82b6e5572e0 change-id: 20231005-ftrace_support-6869d4156139 Best regards,
[PATCH v4] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever we added trace events removing debug messages. Signed-off-by: Krishna chaitanya chundru --- Changes in v4: - Fix compilation issues in previous patch which happended due to rebasing. - In the defconfig FTRACE config is not enabled due to that the compilation issue is not - seen in my workspace. - Link to v3: https://lore.kernel.org/r/2023-ftrace_support-v3-1-f358d2911...@quicinc.com Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 24 +++-- drivers/bus/mhi/host/pm.c | 6 +- drivers/bus/mhi/host/trace.h| 225 5 files changed, 246 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a02a71605907 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include "trace.h" #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..923d51904f35 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -491,11 +491,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,10 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_ctrl_ev_ring(mhi_cntrl->mhi_dev->name, local_rp, + local_rp->ptr, local_rp->dword[0], + local_rp->dword[1]); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +999,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_data_event_ring(mhi_cntrl->mhi_dev->name, local_rp->ptr, + local_rp->dword[0], local_rp->dword[1]); + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1240,8 @@ int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan, mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(info->len); mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, eot, eob, chain); + trace_mhi_gen_tre(mhi_cntrl->mhi_
[PATCH v3] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_mhi_states 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Where ever we added trace events removing debug messages. Signed-off-by: Krishna chaitanya chundru --- Changes in v3: - move trace header file from include/trace/events to drivers/bus/mhi/host/ so that - we can include driver header files. - Use macros directly in the trace events as suggested Jeffrey Hugo. - Reorder the structure in the events as suggested by steve to avoid holes in the buffer. - removed the mhi_to_physical function as this can give security issues. - removed macros to define strings as we can get those from driver headers. - Link to v2: https://lore.kernel.org/r/20231013-ftrace_support-v2-1-6e893ce01...@quicinc.com Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 24 +++-- drivers/bus/mhi/host/pm.c | 6 +- drivers/bus/mhi/host/trace.h| 224 5 files changed, 245 insertions(+), 13 deletions(-) diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..6acb85f4c5f8 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include "trace.h" + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a02a71605907 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include "trace.h" #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..7fe8b2dd9e24 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -491,11 +491,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_states(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +830,10 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_ctrl_ev_ring(mhi_cntrl->mhi_dev->name, (u64)local_rp, + local_rp->ptr, local_rp->dword[0], + local_rp->dword[1]); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +999,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_data_event_ring(mhi_cntrl->mhi_dev->name, local_rp->ptr, + local_rp->dword[0], local_rp->dword[1]); + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1240,8 @@ int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan, mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(info->len); mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, eot, eob, chain); + trace_mhi_gen_tre(mhi_cntrl->mhi_dev->name, mhi_chan->chan, (u64)mhi_tre, + mhi_tre->ptr, mhi_tre->dword[0], mhi_tre->dword[1]); /* increment WP */ mhi_add_ring_element(mhi_cntrl, tre_ring); mhi_add_ring_element(mhi_cntrl, buf_ring); @@ -1327,9 +1334,8 @@ sta
Re: [PATCH v2] bus: mhi: host: Add tracing support
On 10/27/2023 8:59 PM, Jeffrey Hugo wrote: On 10/23/2023 1:11 AM, Krishna Chaitanya Chundru wrote: On 10/20/2023 8:33 PM, Jeffrey Hugo wrote: On 10/13/2023 3:52 AM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_threaded_handler 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Signed-off-by: Krishna chaitanya chundru --- Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- MAINTAINERS | 1 + drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 32 +++-- drivers/bus/mhi/host/pm.c | 6 +- include/trace/events/mhi_host.h | 287 6 files changed, 317 insertions(+), 13 deletions(-) diff --git a/MAINTAINERS b/MAINTAINERS index 35977b269d5e..4339c668a6ab 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -13862,6 +13862,7 @@ F: Documentation/mhi/ F: drivers/bus/mhi/ F: drivers/pci/endpoint/functions/pci-epf-mhi.c F: include/linux/mhi.h +F: include/trace/events/mhi_host.h MICROBLAZE ARCHITECTURE M: Michal Simek diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..3afa90a204fd 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include This feels redundant to me. A few lines ago we included internal.h, and internal.h includes trace/events/mhi_host.h As Steve mentioned, this is mandatory step for creating trace points & trace events. I understand this creates the trace points, and that needs to be done in C code. It dtill seems redundant because we are including the header twice (and I am aware trace has the special multi-header read functionality for this). The duplicate include still feels weird, but I have not come up with a better way to structure this. We will use this way for now then, we will check in parallel if there is a way to avoid this and change it in the future. + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a80a317a59a9 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..fcdb728ba49f 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -246,6 +246,11 @@ static void *mhi_to_virtual(struct mhi_ring *ring, dma_addr_t addr) return (addr - ring->iommu_base) + ring->base; } +dma_addr_t mhi_to_physical(struct mhi_ring *ring, void *addr) +{ + return (addr - ring->base) + ring->iommu_base; +} This seems to be poorly named since we are using the iommu_base which suggests we are converting to an IOVA. Why do we need this though? This seems like it might be a security issue, or at the very least, not preferred, and I'm struggling to figure out what value this provides to you are I when looking at the log. I will rename the function to reflect it is converting to IOVA. We MHI TRE we write the IOVA address, to correlate between TRE events in the MHI ring and event we are processing we want to log the IOVA address. As we are logging only IOVA address which is provided in the devicetree and not the original physical address we are not expecting any security issues here. Correct me if I was wrong. The IOVA is not provided by DT, it is a runtime allocated value provided by the IOMMU, if present. If not present, then it is a physical address. Remember, x86 does not use devicetree. While the IOVA (with an iommu) is not technically a physical address, but is treated as such by the device. I can imagine an attacker doing bad things if they get a hold of the value. Sure we will remove it. Still, you haven't indicated why this is useful. The TRE ring elements has address in the IOVA format when we want to correlate the address with the TRE elements in the dumps it will easier with this way. Anyway we will not expose this as you suggested as it might expose physical address in some platforms. +
Re: [PATCH v2] bus: mhi: host: Add tracing support
On 10/20/2023 8:33 PM, Jeffrey Hugo wrote: On 10/13/2023 3:52 AM, Krishna chaitanya chundru wrote: This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_threaded_handler 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Signed-off-by: Krishna chaitanya chundru --- Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- MAINTAINERS | 1 + drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 32 +++-- drivers/bus/mhi/host/pm.c | 6 +- include/trace/events/mhi_host.h | 287 6 files changed, 317 insertions(+), 13 deletions(-) diff --git a/MAINTAINERS b/MAINTAINERS index 35977b269d5e..4339c668a6ab 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -13862,6 +13862,7 @@ F: Documentation/mhi/ F: drivers/bus/mhi/ F: drivers/pci/endpoint/functions/pci-epf-mhi.c F: include/linux/mhi.h +F: include/trace/events/mhi_host.h MICROBLAZE ARCHITECTURE M: Michal Simek diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..3afa90a204fd 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include This feels redundant to me. A few lines ago we included internal.h, and internal.h includes trace/events/mhi_host.h As Steve mentioned, this is mandatory step for creating trace points & trace events. + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a80a317a59a9 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..fcdb728ba49f 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -246,6 +246,11 @@ static void *mhi_to_virtual(struct mhi_ring *ring, dma_addr_t addr) return (addr - ring->iommu_base) + ring->base; } +dma_addr_t mhi_to_physical(struct mhi_ring *ring, void *addr) +{ + return (addr - ring->base) + ring->iommu_base; +} This seems to be poorly named since we are using the iommu_base which suggests we are converting to an IOVA. Why do we need this though? This seems like it might be a security issue, or at the very least, not preferred, and I'm struggling to figure out what value this provides to you are I when looking at the log. I will rename the function to reflect it is converting to IOVA. We MHI TRE we write the IOVA address, to correlate between TRE events in the MHI ring and event we are processing we want to log the IOVA address. As we are logging only IOVA address which is provided in the devicetree and not the original physical address we are not expecting any security issues here. Correct me if I was wrong. + static void mhi_add_ring_element(struct mhi_controller *mhi_cntrl, struct mhi_ring *ring) { @@ -491,11 +496,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_threaded_handler(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); Why are we removing the debug message when adding this trace? The commit text doesn't say. (Looks like you do this several times, assume this comment applies to all isntances) I will add this in the commit text in my next patch. Just a query is recommended to keep both debug message and trace events. If yes we will not remove the debug messages. if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +835,12 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp
Re: [PATCH v2] bus: mhi: host: Add tracing support
On 10/16/2023 8:43 PM, Steven Rostedt wrote: On Fri, 13 Oct 2023 15:22:19 +0530 Krishna chaitanya chundru wrote: +++ b/include/trace/events/mhi_host.h @@ -0,0 +1,287 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Copyright (c) 2023 Qualcomm Innovation Center, Inc. All rights reserved. + */ + +#undef TRACE_SYSTEM +#define TRACE_SYSTEM mhi_host + +#if !defined(_TRACE_EVENT_MHI_HOST_H) || defined(TRACE_HEADER_MULTI_READ) +#define _TRACE_EVENT_MHI_HOST_H + +#include +#define MHI_STATE \ + EM(MHI_STATE_RESET, "RESET") \ + EM(MHI_STATE_READY, "READY") \ + EM(MHI_STATE_M0,"M0") \ + EM(MHI_STATE_M1,"M1") \ + EM(MHI_STATE_M2,"M2") \ + EM(MHI_STATE_M3,"M3") \ + EM(MHI_STATE_M3_FAST, "M3 FAST")\ + EM(MHI_STATE_BHI, "BHI")\ + EMe(MHI_STATE_SYS_ERR, "SYS ERROR") + +#define MHI_EE \ + EM(MHI_EE_PBL, "PRIMARY BOOTLOADER") \ + EM(MHI_EE_SBL, "SECONDARY BOOTLOADER") \ + EM(MHI_EE_AMSS, "MISSION MODE") \ + EM(MHI_EE_RDDM, "RAMDUMP DOWNLOAD MODE") \ + EM(MHI_EE_WFW, "WLAN FIRMWARE") \ + EM(MHI_EE_PTHRU,"PASS THROUGH") \ + EM(MHI_EE_EDL, "EMERGENCY DOWNLOAD") \ + EM(MHI_EE_FP, "FLASH PROGRAMMER") \ + EM(MHI_EE_DISABLE_TRANSITION, "DISABLE")\ + EMe(MHI_EE_NOT_SUPPORTED, "NOT SUPPORTED") + +#define MHI_PM_STATE \ + EM(MHI_PM_STATE_DISABLE,"DISABLE")\ + EM(MHI_PM_STATE_POR,"POWER ON RESET") \ + EM(MHI_PM_STATE_M0, "M0") \ + EM(MHI_PM_STATE_M2, "M2") \ + EM(MHI_PM_STATE_M3_ENTER, "M?->M3") \ + EM(MHI_PM_STATE_M3, "M3") \ + EM(MHI_PM_STATE_M3_EXIT,"M3->M0") \ + EM(MHI_PM_STATE_FW_DL_ERR, "Firmware Download Error")\ + EM(MHI_PM_STATE_SYS_ERR_DETECT, "SYS ERROR Detect") \ + EM(MHI_PM_STATE_SYS_ERR_PROCESS,"SYS ERROR Process") \ + EM(MHI_PM_STATE_SHUTDOWN_PROCESS, "SHUTDOWN Process") \ + EMe(MHI_PM_STATE_LD_ERR_FATAL_DETECT, "Linkdown or Error Fatal Detect") + +#define MHI_CH_STATE \ + EM(MHI_CH_STATE_TYPE_RESET, "RESET") \ + EM(MHI_CH_STATE_TYPE_STOP, "STOP")\ + EMe(MHI_CH_STATE_TYPE_START, "START") + +#define MHI_DEV_ST_TRANSITION \ + EM(DEV_ST_TRANSITION_PBL, "PBL")\ + EM(DEV_ST_TRANSITION_READY, "READY") \ + EM(DEV_ST_TRANSITION_SBL, "SBL")\ + EM(DEV_ST_TRANSITION_MISSION_MODE, "MISSION MODE") \ + EM(DEV_ST_TRANSITION_FP,"FLASH PROGRAMMER") \ + EM(DEV_ST_TRANSITION_SYS_ERR, "SYS ERROR") \ + EMe(DEV_ST_TRANSITION_DISABLE, "DISABLE") + +/* Enums require being exported to userspace, for user tool parsing */ +#undef EM +#undef EMe +#defineEM(a, b)TRACE_DEFINE_ENUM(a); +#defineEMe(a, b) TRACE_DEFINE_ENUM(a); + +MHI_STATE +MHI_EE +MHI_PM_STATE +MHI_CH_STATE +MHI_DEV_ST_TRANSITION + +/* + * Now redefine the EM() and EMe() macros to map the enums to the strings + * that will be printed in the output. + */ +#undef EM +#undef EMe +#define EM(a, b) {a, b}, +#define EMe(a, b) {a, b} + +TRACE_EVENT(mhi_gen_tre, + + TP_PROTO(const char *name, int ch_num, u64 wp, __le64 tre_ptr, +__le32 dword0, __le32 dword1), + + TP_ARGS(name, ch_num, wp, tre_ptr, dword0, dword1), + + TP_STRUCT__entry( + __string(name, name) + __field(int, ch_num) This is fine as __string() is four bytes in the event (2 bytes for offset where the string exists, and 2 bytes for its length). + __field(u64,
Re: [PATCH] bus: mhi: host: Add tracing support
On 10/6/2023 4:10 AM, Bjorn Andersson wrote: On Thu, Oct 05, 2023 at 03:55:20PM +0530, Krishna chaitanya chundru wrote: This change adds ftrace support for following: 1. mhi_intvec_threaded_handler 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker This is not the best "problem description". Usage: echo 1 > /sys/kernel/debug/tracing/events/mhi_host/enable cat /sys/kernel/debug/tracing/trace This does not need to be included in the commit message, how to use the tracing framework is documented elsewhere. Removed this now. [..] diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..499590437e9b 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -491,11 +491,10 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_threaded_handler(mhi_cntrl->mhi_dev->name, TO_MHI_EXEC_STR(mhi_cntrl->ee), + mhi_state_str(mhi_cntrl->dev_state), + TO_MHI_EXEC_STR(ee), mhi_state_str(state)); All these helper functions that translates a state to a string, pass the raw state into the trace event and use __print_symbolic() in your TP_printk() instead. This will allow you to read the state, but you can have tools act of the numerical value. (This comment applies to all the trace events) changed the events as you suggested. if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, [..] diff --git a/include/trace/events/mhi_host.h b/include/trace/events/mhi_host.h [..] + +TRACE_EVENT(mhi_pm_st_worker, Why is this trace event called "worker", isn't the event a "mhi_pm_state_transition"? Don't just name your trace event based on the function that triggers them, but what they represent and make sure they carry useful information to understand the system. If you want to trace the flow through your functions, you can use e.g. ftrace. Regards, Bjorn Changed as you suggested. - KC
[PATCH v2] bus: mhi: host: Add tracing support
This change adds ftrace support for following functions which helps in debugging the issues when there is Channel state & MHI state change and also when we receive data and control events: 1. mhi_intvec_threaded_handler 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Signed-off-by: Krishna chaitanya chundru --- Changes in v2: - Passing the raw state into the trace event and using __print_symbolic() as suggested by bjorn. - Change mhi_pm_st_worker to mhi_pm_st_transition as suggested by bjorn. - Fixed the kernel test rebot issues. - Link to v1: https://lore.kernel.org/r/20231005-ftrace_support-v1-1-23a2f394f...@quicinc.com --- MAINTAINERS | 1 + drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 32 +++-- drivers/bus/mhi/host/pm.c | 6 +- include/trace/events/mhi_host.h | 287 6 files changed, 317 insertions(+), 13 deletions(-) diff --git a/MAINTAINERS b/MAINTAINERS index 35977b269d5e..4339c668a6ab 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -13862,6 +13862,7 @@ F: Documentation/mhi/ F: drivers/bus/mhi/ F: drivers/pci/endpoint/functions/pci-epf-mhi.c F: include/linux/mhi.h +F: include/trace/events/mhi_host.h MICROBLAZE ARCHITECTURE M: Michal Simek diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..3afa90a204fd 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a80a317a59a9 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..fcdb728ba49f 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -246,6 +246,11 @@ static void *mhi_to_virtual(struct mhi_ring *ring, dma_addr_t addr) return (addr - ring->iommu_base) + ring->base; } +dma_addr_t mhi_to_physical(struct mhi_ring *ring, void *addr) +{ + return (addr - ring->base) + ring->iommu_base; +} + static void mhi_add_ring_element(struct mhi_controller *mhi_cntrl, struct mhi_ring *ring) { @@ -491,11 +496,9 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_threaded_handler(mhi_cntrl->mhi_dev->name, mhi_cntrl->ee, + mhi_cntrl->dev_state, ee, state); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +835,12 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_ctrl_ev_ring(mhi_cntrl->mhi_dev->name, + mhi_to_physical(ev_ring, local_rp), + local_rp->ptr, local_rp->dword[0], + local_rp->dword[1], + MHI_TRE_GET_EV_STATE(local_rp)); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +1006,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_data_event_ring(mhi_cntrl->mhi_dev->name, local_rp->ptr, + local_rp->dword[0], local_rp->dword[1]); + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1247,9 @@ int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan, mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(info->len); mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, eot, eob
[PATCH] bus: mhi: host: Add tracing support
This change adds ftrace support for following: 1. mhi_intvec_threaded_handler 2. mhi_process_data_event_ring 3. mhi_process_ctrl_ev_ring 4. mhi_gen_tre 5. mhi_update_channel_state 6. mhi_tryset_pm_state 7. mhi_pm_st_worker Usage: echo 1 > /sys/kernel/debug/tracing/events/mhi_host/enable cat /sys/kernel/debug/tracing/trace Signed-off-by: Krishna chaitanya chundru --- MAINTAINERS | 1 + drivers/bus/mhi/host/init.c | 3 + drivers/bus/mhi/host/internal.h | 1 + drivers/bus/mhi/host/main.c | 27 -- drivers/bus/mhi/host/pm.c | 7 +- include/trace/events/mhi_host.h | 207 6 files changed, 234 insertions(+), 12 deletions(-) diff --git a/MAINTAINERS b/MAINTAINERS index 35977b269d5e..4339c668a6ab 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -13862,6 +13862,7 @@ F: Documentation/mhi/ F: drivers/bus/mhi/ F: drivers/pci/endpoint/functions/pci-epf-mhi.c F: include/linux/mhi.h +F: include/trace/events/mhi_host.h MICROBLAZE ARCHITECTURE M: Michal Simek diff --git a/drivers/bus/mhi/host/init.c b/drivers/bus/mhi/host/init.c index f78aefd2d7a3..3afa90a204fd 100644 --- a/drivers/bus/mhi/host/init.c +++ b/drivers/bus/mhi/host/init.c @@ -20,6 +20,9 @@ #include #include "internal.h" +#define CREATE_TRACE_POINTS +#include + static DEFINE_IDA(mhi_controller_ida); const char * const mhi_ee_str[MHI_EE_MAX] = { diff --git a/drivers/bus/mhi/host/internal.h b/drivers/bus/mhi/host/internal.h index 2e139e76de4c..a80a317a59a9 100644 --- a/drivers/bus/mhi/host/internal.h +++ b/drivers/bus/mhi/host/internal.h @@ -7,6 +7,7 @@ #ifndef _MHI_INT_H #define _MHI_INT_H +#include #include "../common.h" extern struct bus_type mhi_bus_type; diff --git a/drivers/bus/mhi/host/main.c b/drivers/bus/mhi/host/main.c index dcf627b36e82..499590437e9b 100644 --- a/drivers/bus/mhi/host/main.c +++ b/drivers/bus/mhi/host/main.c @@ -491,11 +491,10 @@ irqreturn_t mhi_intvec_threaded_handler(int irq_number, void *priv) state = mhi_get_mhi_state(mhi_cntrl); ee = mhi_get_exec_env(mhi_cntrl); - dev_dbg(dev, "local ee: %s state: %s device ee: %s state: %s\n", - TO_MHI_EXEC_STR(mhi_cntrl->ee), - mhi_state_str(mhi_cntrl->dev_state), - TO_MHI_EXEC_STR(ee), mhi_state_str(state)); + trace_mhi_intvec_threaded_handler(mhi_cntrl->mhi_dev->name, TO_MHI_EXEC_STR(mhi_cntrl->ee), + mhi_state_str(mhi_cntrl->dev_state), + TO_MHI_EXEC_STR(ee), mhi_state_str(state)); if (state == MHI_STATE_SYS_ERR) { dev_dbg(dev, "System error detected\n"); pm_state = mhi_tryset_pm_state(mhi_cntrl, @@ -832,6 +831,11 @@ int mhi_process_ctrl_ev_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_ctrl_ev_ring(mhi_cntrl->mhi_dev->name, (u64)(local_rp), + local_rp->ptr, local_rp->dword[0], + local_rp->dword[1], + mhi_state_str(MHI_TRE_GET_EV_STATE(local_rp))); + switch (type) { case MHI_PKT_TYPE_BW_REQ_EVENT: { @@ -997,6 +1001,9 @@ int mhi_process_data_event_ring(struct mhi_controller *mhi_cntrl, while (dev_rp != local_rp && event_quota > 0) { enum mhi_pkt_type type = MHI_TRE_GET_EV_TYPE(local_rp); + trace_mhi_process_data_event_ring(mhi_cntrl->mhi_dev->name, local_rp->ptr, + local_rp->dword[0], local_rp->dword[1]); + chan = MHI_TRE_GET_EV_CHID(local_rp); WARN_ON(chan >= mhi_cntrl->max_chan); @@ -1235,6 +1242,8 @@ int mhi_gen_tre(struct mhi_controller *mhi_cntrl, struct mhi_chan *mhi_chan, mhi_tre->dword[0] = MHI_TRE_DATA_DWORD0(info->len); mhi_tre->dword[1] = MHI_TRE_DATA_DWORD1(bei, eot, eob, chain); + trace_mhi_gen_tre(mhi_cntrl->mhi_dev->name, mhi_chan->chan, (u64)(mhi_tre), + mhi_tre->ptr, mhi_tre->dword[0], mhi_tre->dword[1]); /* increment WP */ mhi_add_ring_element(mhi_cntrl, tre_ring); mhi_add_ring_element(mhi_cntrl, buf_ring); @@ -1327,9 +1336,8 @@ static int mhi_update_channel_state(struct mhi_controller *mhi_cntrl, enum mhi_cmd_type cmd = MHI_CMD_NOP; int ret; - dev_dbg(dev, "%d: Updating channel state to: %s\n", mhi_chan->chan, - TO_CH_STATE_TYPE_STR(to_state)); - + trace_mhi_update_channel_state_start(mhi_cntrl->mhi_dev->name, mhi_