arm64: Convert to ARCH_STACKWALK
Subsystems (6)
v2 cover arm64: Convert to ARCH_STACKWALK
This series updates the arm64 stacktrace code to use the newer and much simpler arch_stack_walk() interface, the main benefit being a single entry point to the arch code with no need for the arch code to worry about skipping frames. Along the way I noticed that the reliable parameter to the arch_stack_walk() callback appears to be redundant so there's also a patch here removing that from the existing code to simplify the interface.
This is preparatory work for implementing reliable stack trace for arm64.
v2: Rebase onto v5.9-rc1.
Mark Brown (3):
stacktrace: Remove reliable argument from arch_stack_walk() callback
arm64: stacktrace: Make stack walk callback consistent with generic
code
arm64: stacktrace: Convert to ARCH_STACKWALK
arch/arm64/Kconfig | 1 + arch/arm64/include/asm/stacktrace.h | 2 +- arch/arm64/kernel/perf_callchain.c | 6 +-- arch/arm64/kernel/return_address.c | 8 +-- arch/arm64/kernel/stacktrace.c | 84 ++++------------------------- arch/s390/kernel/stacktrace.c | 4 +- arch/x86/kernel/stacktrace.c | 10 ++-- include/linux/stacktrace.h | 5 +- kernel/stacktrace.c | 8 ++- 9 files changed, 30 insertions(+), 98 deletions(-)
Signature · 1 line
2.20.1
PATCH v2 1/3 stacktrace: Remove reliable argument from arch_stack_walk() callback
Currently the callback passed to arch_stack_walk() has an argument called reliable passed to it to indicate if the stack entry is reliable, a comment says that this is used by some printk() consumers. However in the current kernel none of the arch_stack_walk() implementations ever set this flag to true and the only callback implementation we have is in the generic stacktrace code which ignores the flag. It therefore appears that this flag is redundant so we can simplify and clarify things by removing it.
--- arch/s390/kernel/stacktrace.c | 4 ++-- arch/x86/kernel/stacktrace.c | 10 +++++----- include/linux/stacktrace.h | 5 +---- kernel/stacktrace.c | 8 +++----- 4 files changed, 11 insertions(+), 16 deletions(-)
arch/s390/kernel/stacktrace.c +2 -2
@@ -19,7 +19,7 @@ void arch_stack_walk(stack_trace_consume_fn consume_entry, void *cookie,unwind_for_each_frame (&state, task, regs,0 ) { addr =unwind_get_return_address (&state); -if (!addr || !consume_entry (cookie, addr, false)) +if (!addr || !consume_entry (cookie, addr))break ; } }
@@ -56,7 +56,7 @@ int arch_stack_walk_reliable(stack_trace_consume_fn consume_entry,return -EINVAL ;#endif -if (!consume_entry (cookie, addr, false)) +if (!consume_entry (cookie, addr))return -EINVAL ; }
arch/x86/kernel/stacktrace.c +5 -5
@@ -18,13 +18,13 @@ void arch_stack_walk(stack_trace_consume_fn consume_entry, void *cookie,struct unwind_state state;unsigned long addr; -if (regs && !consume_entry (cookie, regs->ip, false)) +if (regs && !consume_entry (cookie, regs->ip))return ;for (unwind_start (&state, task, regs,NULL ); !unwind_done (&state);unwind_next_frame (&state)) { addr =unwind_get_return_address (&state); -if (!addr || !consume_entry (cookie, addr, false)) +if (!addr || !consume_entry (cookie, addr))break ; } }
@@ -72,7 +72,7 @@ int arch_stack_walk_reliable(stack_trace_consume_fn consume_entry,if (!addr)return -EINVAL ; -if (!consume_entry (cookie, addr, false)) +if (!consume_entry (cookie, addr))return -EINVAL ; }
@@ -114,7 +114,7 @@ void arch_stack_walk_user(stack_trace_consume_fn consume_entry, void *cookie, {const void __user *fp = (const void __user *)regs->bp; -if (!consume_entry (cookie, regs->ip, false)) +if (!consume_entry (cookie, regs->ip))return ;while (1 ) {
@@ -128,7 +128,7 @@ void arch_stack_walk_user(stack_trace_consume_fn consume_entry, void *cookie,break ;if (!frame.ret_addr)break ; -if (!consume_entry (cookie, frame.ret_addr, false)) +if (!consume_entry (cookie, frame.ret_addr))break ; fp = frame.next_fp; }
include/linux/stacktrace.h +1 -4
@@ -29,14 +29,11 @@ unsigned int stack_trace_save_user(unsigned long *store, unsigned int size);* stack_trace_consume_fn - Callback for arch_stack_walk() * @cookie: Caller supplied pointer handed back by arch_stack_walk() * @addr: The stack entry address to consume -* @reliable: True when the stack entry is reliable. Required by -* some printk based consumers. * * Return: True, if the entry was consumed or skipped * False, if there is no space left to store */ -typedef bool (*stack_trace_consume_fn )(void *cookie,unsigned long addr, -bool reliable); +typedef bool (*stack_trace_consume_fn )(void *cookie,unsigned long addr);/** * arch_stack_walk - Architecture specific function to walk the stack * @consume_entry: Callback which is invoked by the architecture code for
kernel/stacktrace.c +3 -5
@@ -78,8 +78,7 @@ struct stacktrace_cookie {unsigned int len; }; -static bool stack_trace_consume_entry (void *cookie,unsigned long addr, -bool reliable) +static bool stack_trace_consume_entry (void *cookie,unsigned long addr) {struct stacktrace_cookie *c = cookie;
@@ -94,12 +93,11 @@ static bool stack_trace_consume_entry(void *cookie, unsigned long addr,return c->len < c->size; } -static bool stack_trace_consume_entry_nosched (void *cookie,unsigned long addr, -bool reliable) +static bool stack_trace_consume_entry_nosched (void *cookie,unsigned long addr) {if (in_sched_functions (addr))return true; -return stack_trace_consume_entry (cookie, addr, reliable); +return stack_trace_consume_entry (cookie, addr); }/**
Signature · 1 line
2.20.1
PATCH v2 2/3 arm64: stacktrace: Make stack walk callback consistent with generic code
In reply to Mark BrownAs with the generic arch_stack_walk() code the arm64 stack walk code takes a callback that is called per stack frame. Currently the arm64 code always passes a struct stackframe to the callback and the generic code just passes the pc, however none of the users ever reference anything in the struct other than the pc value. The arm64 code also uses a return type of int while the generic code uses a return type of bool though in both cases the return value is a boolean value.
In order to reduce code duplication when arm64 is converted to use arch_stack_walk() change the signature of the arm64 specific callback to match that of the generic code.
--- arch/arm64/include/asm/stacktrace.h | 2 +- arch/arm64/kernel/perf_callchain.c | 6 +++--- arch/arm64/kernel/return_address.c | 8 ++++---- arch/arm64/kernel/stacktrace.c | 11 +++++------ 4 files changed, 13 insertions(+), 14 deletions(-)
arch/arm64/include/asm/stacktrace.h +1 -1
@@ -63,7 +63,7 @@ struct stackframe {extern int unwind_frame (struct task_struct *tsk,struct stackframe *frame);extern void walk_stackframe (struct task_struct *tsk,struct stackframe *frame, -int (*fn)(struct stackframe *,void *),void *data); +bool (*fn)(void *,unsigned long ),void *data);extern void dump_backtrace (struct pt_regs *regs,struct task_struct *tsk,const char *loglvl);
arch/arm64/kernel/perf_callchain.c +3 -3
@@ -137,11 +137,11 @@ void perf_callchain_user(struct perf_callchain_entry_ctx *entry,* whist unwinding the stackframe and is like a subroutine return so we use * the PC. */ -static int callchain_trace (struct stackframe *frame,void *data) +static bool callchain_trace (void *data,unsigned long pc) {struct perf_callchain_entry_ctx *entry = data; -perf_callchain_store (entry, frame->pc); -return 0 ; +perf_callchain_store (entry, pc); +return false; } void perf_callchain_kernel(struct perf_callchain_entry_ctx *entry,
arch/arm64/kernel/return_address.c +4 -4
@@ -18,16 +18,16 @@ struct return_address_data {void *addr; }; -static int save_return_addr (struct stackframe *frame,void *d) +static bool save_return_addr (void *d,unsigned long pc) {struct return_address_data *data = d;if (!data->level) { - data->addr = (void *)frame->pc; -return 1 ; + data->addr = (void *)pc; +return true; }else { --data->level; -return 0 ; +return false; } }NOKPROBE_SYMBOL (save_return_addr);
arch/arm64/kernel/stacktrace.c +5 -6
@@ -118,12 +118,12 @@ int notrace unwind_frame(struct task_struct *tsk, struct stackframe *frame)NOKPROBE_SYMBOL (unwind_frame); void notrace walk_stackframe(struct task_struct *tsk, struct stackframe *frame, -int (*fn)(struct stackframe *,void *),void *data) +bool (*fn)(void *,unsigned long ),void *data) {while (1 ) {int ret; -if (fn (frame, data)) +if (fn (data, frame->pc))break ; ret =unwind_frame (tsk, frame);if (ret <0 )
@@ -139,17 +139,16 @@ struct stack_trace_data {unsigned int skip; }; -static int save_trace (struct stackframe *frame,void *d) +static bool save_trace (void *d,unsigned long addr) {struct stack_trace_data *data = d;struct stack_trace *trace = data->trace; -unsigned long addr = frame->pc;if (data->no_sched_functions &&in_sched_functions (addr)) -return 0 ; +return false;if (data->skip) { data->skip--; -return 0 ; +return false; } trace->entries[trace->nr_entries++] = addr;
Signature · 1 line
2.20.1
PATCH v2 3/3 arm64: stacktrace: Convert to ARCH_STACKWALK
In reply to Mark BrownHistorically architectures have had duplicated code in their stack trace implementations for filtering what gets traced. In order to avoid this duplication some generic code has been provided using a new interface arch_stack_walk(), enabled by selecting ARCH_STACKWALK in Kconfig, which factors all this out into the generic stack trace code. Convert arm64 to use this common infrastructure.
--- arch/arm64/Kconfig | 1 + arch/arm64/kernel/stacktrace.c | 79 ++++------------------------------ 2 files changed, 9 insertions(+), 71 deletions(-)
arch/arm64/Kconfig +1 -0
@@ -29,6 +29,7 @@ config ARM64 select ARCH_HAS_SETUP_DMA_OPS select ARCH_HAS_SET_DIRECT_MAP select ARCH_HAS_SET_MEMORY + select ARCH_STACKWALK select ARCH_HAS_STRICT_KERNEL_RWX select ARCH_HAS_STRICT_MODULE_RWX select ARCH_HAS_SYNC_DMA_FOR_DEVICE
arch/arm64/kernel/stacktrace.c +8 -71
90 lines not on this page · raw message
Signature · 1 line
2.20.1
Hi,
Quoted 10 lines · Mark Brown
This series updates the arm64 stacktrace code to use the newer and much simpler
arch_stack_walk()interface, the main benefit being a single entry point to the arch code with no need for the arch code to worry about skipping frames. Along the way I noticed that the reliable parameter to thearch_stack_walk()callback appears to be redundant so there's also a patch here removing that from the existing code to simplify the interface.This is preparatory work for implementing reliable stack trace for arm64.
This all looks pretty nice!
Just to check, has the skipping logic been tested to work equivalently to what we had before? By inspection I think it should, but since it relies on function call boundaries it always strikes me as fragile.
If you could confirm that (e.g. with LKDTM perhaps?) that'd be great. Assuming that looks right, for the series:
Mark.
Quoted 23 lines
v2: Rebase onto v5.9-rc1.
Mark Brown (3):
stacktrace: Remove reliable argument fromarch_stack_walk()callback
arm64: stacktrace: Make stack walk callback consistent with generic
code
arm64: stacktrace: Convert toARCH_STACKWALKarch/arm64/Kconfig | 1 + arch/arm64/include/asm/stacktrace.h | 2 +- arch/arm64/kernel/perf_callchain.c | 6 +-- arch/arm64/kernel/return_address.c | 8 +-- arch/arm64/kernel/stacktrace.c | 84 ++++------------------------- arch/s390/kernel/stacktrace.c | 4 +- arch/x86/kernel/stacktrace.c | 10 ++-- include/linux/stacktrace.h | 5 +- kernel/stacktrace.c | 8 ++- 9 files changed, 30 insertions(+), 98 deletions(-)Signature · 1 line
2.20.1
Re: [PATCH v2 1/3] stacktrace: Remove reliable argument from arch_stack_walk() callback
In reply to Mark BrownQuoted 9 lines · Mark Brown
Currently the callback passed to
arch_stack_walk()has an argument called reliable passed to it to indicate if the stack entry is reliable, a comment says that this is used by someprintk()consumers. However in the current kernel none of thearch_stack_walk()implementations ever set this flag to true and the only callback implementation we have is in the generic stacktrace code which ignores the flag. It therefore appears that this flag is redundant so we can simplify and clarify things by removing it.Signed-off-by Mark Brown <[email protected]>
M
Re: [PATCH v2 2/3] arm64: stacktrace: Make stack walk callback consistent with generic code
In reply to Mark BrownQuoted 8 lines · Mark Brown
As with the generic
arch_stack_walk()code the arm64 stack walk code takes a callback that is called per stack frame. Currently the arm64 code always passes a struct stackframe to the callback and the generic code just passes the pc, however none of the users ever reference anything in the struct other than the pc value. The arm64 code also uses a return type of int while the generic code uses a return type of bool though in both cases the return value is a boolean value.
Quoted 5 lines
In order to reduce code duplication when arm64 is converted to use
arch_stack_walk()change the signature of the arm64 specific callback to match that of the generic code.Signed-off-by Mark Brown <[email protected]>
M
Re: [PATCH v2 3/3] arm64: stacktrace: Convert to ARCH_STACKWALK
In reply to Mark BrownHi,
it could be a silly question, but better to ask...
arch/arm64/kernel/stacktrace.c void notrace walk_stackframe(struct task_struct *tsk, struct stackframe *frame, · line 146
Quoted 5 lines
arch/arm64/kernel/stacktrace.c +5
-0+if (regs) +start_backtrace (&frame, regs->regs[29 ], regs->pc); +else +start_backtrace (&frame,thread_saved_fp (task), +thread_saved_pc (task));
Would this also work for task == current? Given that the original code had
arch/arm64/kernel/stacktrace.c void notrace walk_stackframe(struct task_struct *tsk, struct stackframe *frame, · line 194
Quoted 3 lines
arch/arm64/kernel/stacktrace.c +0
-3-start_backtrace (&frame, - (unsigned long )__builtin_frame_address (0 ), - (unsigned long )__save_stack_trace);
for the case, which seems correct (but I don't know much about arm64 arch in the kernel).
Otherwise, I did not spot anything suspicious or wrong.
Regards
Miroslav
Quoted 6 lines · Mark Rutland
Just to check, has the skipping logic been tested to work equivalently to what we had before? By inspection I think it should, but since it relies on function call boundaries it always strikes me as fragile.
If you could confirm that (e.g. with LKDTM perhaps?) that'd be great. Assuming that looks right, for the series:
I've tested this with LKDTM and otherwise and didn't spot any issues (and just did a bit of retesting) but it is a pretty manual process so it's possible I missed something.
Mark Rutland
maintainerRe: [PATCH v2 3/3] arm64: stacktrace: Convert to ARCH_STACKWALK
In reply to Miroslav BenesQuoted 12 lines · Miroslav Benes
Hi,
it could be a silly question, but better to ask...
arch/arm64/kernel/stacktrace.c void notrace walk_stackframe(struct task_struct *tsk, struct stackframe *frame, · line 146
Quoted 5 lines
arch/arm64/kernel/stacktrace.c +5
-0+if (regs) +start_backtrace (&frame, regs->regs[29 ], regs->pc); +else +start_backtrace (&frame,thread_saved_fp (task), +thread_saved_pc (task));Would this also work for task == current? Given that the original code had
Quoted 3 lines
arch/arm64/kernel/stacktrace.c void notrace walk_stackframe(struct task_struct *tsk, struct stackframe *frame, · line 194
Quoted 3 lines
arch/arm64/kernel/stacktrace.c +0
-3-start_backtrace (&frame, - (unsigned long )__builtin_frame_address (0 ), - (unsigned long )__save_stack_trace);
Oh whoops; I'm annoyed I didn't spot that.
With that gone this cannot work for (task == current && regs == NULL), as we'll erroneously use stale values from the task struct.
It looks like the LKDTM tests only trigger cases with non-NULL regs, but IIUC this should show up with show_stack(NULL, NULL, KERN_INFO), as drivers/tty/sysrq.c does for other cpus.
Thanks,
Mark.
Mark Rutland
maintainerQuoted 6 lines · Mark Brown
Quoted 3 lines · Mark Rutland
Just to check, has the skipping logic been tested to work equivalently to what we had before? By inspection I think it should, but since it relies on function call boundaries it always strikes me as fragile.
Quoted 6 lines
Quoted 2 lines
If you could confirm that (e.g. with LKDTM perhaps?) that'd be great. Assuming that looks right, for the series:
I've tested this with LKDTM and otherwise and didn't spot any issues (and just did a bit of retesting) but it is a pretty manual process so it's possible I missed something.
It looks like the case Mirolav pointed out (self-bactrace with NULL regs) isn't triggered by LKDTM (since it always causes a backtrace from an exception with non-NULL regs), but we do rely on that elsewhere in the kernel.
It might be worth adding an LKDTM test to trigger that.
I'll wait for a respin that handles that -- please drop the ack for now.
Thanks,
Mark.
Re: [PATCH v2 3/3] arm64: stacktrace: Convert to ARCH_STACKWALK
In reply to Mark RutlandQuoted 10 lines · Mark Rutland
Quoted 3 lines · Miroslav Benes
arch/arm64/kernel/stacktrace.c void notrace walk_stackframe(struct task_struct *tsk, struct stackframe *frame, · line 194
Quoted 3 lines
arch/arm64/kernel/stacktrace.c +0
-3-start_backtrace (&frame, - (unsigned long )__builtin_frame_address (0 ), - (unsigned long )__save_stack_trace);Oh whoops; I'm annoyed I didn't spot that.
With that gone this cannot work for (task == current && regs == NULL), as we'll erroneously use stale values from the task struct.
I remember somehow convincing myself at the time I originally did this that doing the above was redundant with the new interface but that was quite some time ago and I can't reconstruct my reasoning any more, I'm pretty sure I was just mistaken. I've added it back in, thanks for spotting this.
Quoted 3 lines
It looks like the LKDTM tests only trigger cases with non-NULL regs, but IIUC this should show up with
show_stack(NULL, NULL, KERN_INFO), asdrivers/tty/sysrq.cdoes for other cpus.
show_stack() doesn't go through this bit of the stacktrace code, it goes through dump_backtrace() in traps.c which used the underlying arch specific unwinder directly so is unaffected by arch_stack_walk(). Actually now I look at LKDTM it's ending up using show_stack() mostly if not entirely so my testing with it was not exercising this change as much as might be expected anyway (the modified code was getting hit by other things like /proc/N/stack).