xen: add support for automatic debug key actions in case of crash
v4 cover xen: add support for automatic debug key actions in case of crash
When the host crashes it would sometimes be nice to have additional debug data available which could be produced via debug keys, but halting the server for manual intervention might be impossible due to the need to reboot/kexec rather sooner than later.
Add support for automatic debug key actions in case of crashes which can be activated via boot- or runtime-parameter.
Changes in V4:
- addressed comments (now patch 3)
- added patches 1 and 2
Some further remarks to the new patches added:
Patch 1 adds Arm support for run_in_exception_handler(). Constructing the related bug_frame is unfortunately not as easy as on x86.
I have verified that %c in inline assembly isn't supported by gcc 7 for arm64, so this was the only way I've found to support this feature. In theory it might be possible to add a variable referencing the called function and to discard that variable again when linking, but I'd like to add this more intrusive modification only if really wanted.
There is more potential for unifying struct bug_frame between x86 and Arm, either by:
- using the Arm layout on x86, too (resulting in a grow of the bugframe data for the cases without messages)
- trying to construct the data in C instead of inline assembly, which will need to either keep the x86 assembler BUG_FRAME construction, or to add a few functions issuing the ASSERT/BUG/WARN which would then need to be called from *.S files.
Patch 2 opens up more potential for simplification: in theory there is no need any more to call any key handler with the regs parameter, allowing to use the same prototype for all handlers. The downside would be to have an additional irq frame on the stack for the dump_registers() and the do_debug_key() handlers. In case this is acceptable I'd be happy to send a related cleanup patch.
Juergen Gross (3):
xen/arm: add support for run_in_exception_handler()
xen: enable keyhandlers to work without register set specified
xen: add support for automatic debug key actions in case of crash
docs/misc/xen-command-line.pandoc | 41 +++++++++++++++++ xen/arch/arm/traps.c | 10 ++++- xen/common/kexec.c | 8 ++-- xen/common/keyhandler.c | 73 +++++++++++++++++++++++++++++-- xen/common/shutdown.c | 4 +- xen/drivers/char/console.c | 2 +- xen/drivers/char/ns16550.c | 3 +- xen/include/asm-arm/bug.h | 32 +++++++++----- xen/include/xen/kexec.h | 10 ++++- xen/include/xen/keyhandler.h | 10 +++++ 10 files changed, 169 insertions(+), 24 deletions(-)
Signature · 1 line
2.26.2
PATCH v4 1/3 xen/arm: add support for run_in_exception_handler()
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a bug_frame.
Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via run_in_exception_handler() having to be globally visible.
---
V4:
- new patch
I have verified the created bugframe is correct by inspecting the created binary.
--- xen/arch/arm/traps.c | 10 +++++++++- xen/drivers/char/ns16550.c | 3 ++- xen/include/asm-arm/bug.h | 32 +++++++++++++++++++++----------- 3 files changed, 32 insertions(+), 13 deletions(-)
xen/arch/arm/traps.c +9 -1
@@ -1236,8 +1236,16 @@ int do_bug_frame(const struct cpu_user_regs *regs, vaddr_t pc)if ( !bug )return -ENOENT ; +if ( id == BUGFRAME_run_fn ) + { +void (*fn)(const struct cpu_user_regs *) =bug_ptr (bug); + +fn (regs); +return 0 ; + } +/* WARN, BUG or ASSERT: decode the filename pointer and line number. */ - filename =bug_file (bug); + filename =bug_ptr (bug);if ( !is_kernel (filename) )return -EINVAL ; fixup =strlen (filename);
xen/drivers/char/ns16550.c +2 -1
@@ -192,7 +192,8 @@ static void ns16550_interrupt(/* Safe: ns16550_poll() runs as softirq so not reentrant on a given CPU. */ static DEFINE_PER_CPU(struct serial_port *, poll_port); -static void __ns16550_poll (struct cpu_user_regs *regs) +/* run_in_exception_handler() on Arm requires globally visible symbol. */ +void __ns16550_poll (struct cpu_user_regs *regs) {struct serial_port *port =this_cpu (poll_port);struct ns16550 *uart = port->uart;
xen/include/asm-arm/bug.h +21 -11
@@ -15,34 +15,38 @@struct bug_frame {signed int loc_disp;/* Relative address to the bug address */ -signed int file_disp;/* Relative address to the filename */ +signed int ptr_disp;/* Relative address to the filename or function */ signed int msg_disp;/* Relative address to the predicate (for ASSERT) */ uint16_t line;/* Line number */ uint32_t pad0:16 ;/* Padding for 8-bytes align */ };#define bug_loc (b) ((const void *)(b) + (b)->loc_disp) -#define bug_file (b) ((const void *)(b) + (b)->file_disp); +#define bug_ptr (b) ((const void *)(b) + (b)->ptr_disp);#define bug_line (b) ((b)->line)#define bug_msg (b) ((const char *)(b) + (b)->msg_disp) -#define BUGFRAME_warn 0 -#define BUGFRAME_bug 1 -#define BUGFRAME_assert 2 +#define BUGFRAME_run_fn 0 +#define BUGFRAME_warn 1 +#define BUGFRAME_bug 2 +#define BUGFRAME_assert 3 -#define BUGFRAME_NR 3 +#define BUGFRAME_NR 4/* Many versions of GCC doesn't support the asm %c parameter which would * be preferable to this unpleasantness. We use mergeable string * sections to avoid multiple copies of the string appearing in the * Xen image. */ -#define BUG_FRAME (type, line, file, has_msg, msg) do { \ +#define BUG_FRAME (type, line, ptr, ptr_is_file, has_msg, msg) do { \ BUILD_BUG_ON((line) >> 16); \ BUILD_BUG_ON((type) >= BUGFRAME_NR); \ asm ("1:"BUG_INSTR"\n" \ ".pushsection .rodata.str, \"aMS\", %progbits, 1\n" \ - "2:\t.asciz " __stringify(file) "\n" \ + "2:\n" \ + ".if " #ptr_is_file "\n" \ + "\t.asciz " __stringify(ptr) "\n" \ + ".endif\n" \ "3:\n" \ ".if " #has_msg "\n" \ "\t.asciz " #msg "\n" \
@@ -52,21 +56,27 @@ struct bug_frame {"4:\n" \".p2align 2\n" \".long (1b - 4b)\n" \ + ".if " #ptr_is_file "\n" \ ".long (2b - 4b)\n" \ + ".else\n" \ + ".long (" #ptr " - 4b)\n" \ + ".endif\n" \ ".long (3b - 4b)\n" \ ".hword " __stringify(line) ", 0\n" \ ".popsection"); \ } while (0) -#define WARN () BUG_FRAME(BUGFRAME_warn, __LINE__, __FILE__, 0, "") +#define run_in_exception_handler (fn) BUG_FRAME(BUGFRAME_run_fn, 0, fn, 0, 0, "") + +#define WARN () BUG_FRAME(BUGFRAME_warn, __LINE__, __FILE__, 1, 0, "")#define BUG () do { \ - BUG_FRAME(BUGFRAME_bug, __LINE__, __FILE__, 0, ""); \ + BUG_FRAME(BUGFRAME_bug, __LINE__, __FILE__, 1, 0, ""); \ unreachable(); \ } while (0)#define assert_failed (msg) do { \ - BUG_FRAME(BUGFRAME_assert, __LINE__, __FILE__, 1, msg); \ + BUG_FRAME(BUGFRAME_assert, __LINE__, __FILE__, 1, 1, msg); \ unreachable(); \ } while (0)
Signature · 1 line
2.26.2
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.
Could you extend on this? I don't understand what the relevant difference is, from just looking at the changes.
Quoted 8 lines
Signed-off-by Juergen Gross <[email protected]>---
V4:
- new patchI have verified the created bugframe is correct by inspecting the created binary.
Quoted 6 lines
Signed-off-by Juergen Gross <[email protected]>--- xen/arch/arm/traps.c | 10 +++++++++- xen/drivers/char/ns16550.c | 3 ++- xen/include/asm-arm/bug.h | 32 +++++++++++++++++++++----------- 3 files changed, 32 insertions(+), 13 deletions(-)
Aiui you also need to modify xen.lds.S to cover the new (or really the last renamed) section.
Jan
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Quoted 10 lines · Jan Beulich
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.Could you extend on this? I don't understand what the relevant difference is, from just looking at the changes.
The problem seems to be that a static symbol referenced from the inline asm seems not to silence the error that this static symbol isn't being used. On x86 the bug_frame is constructed using the %c modifier, which is not supported for Arm (at least gcc 7 used in my compile test complained), but seems to be enough for gcc on x86 to not complain.
Quoted 16 lines
Quoted 9 lines
Signed-off-by Juergen Gross <[email protected]>---
V4:
- new patchI have verified the created bugframe is correct by inspecting the created binary.
Quoted 6 lines
Signed-off-by Juergen Gross <[email protected]>--- xen/arch/arm/traps.c | 10 +++++++++- xen/drivers/char/ns16550.c | 3 ++- xen/include/asm-arm/bug.h | 32 +++++++++++++++++++++----------- 3 files changed, 32 insertions(+), 13 deletions(-)
Quoted 2 lines
Aiui you also need to modify
xen.lds.Sto cover the new (or really the last renamed) section.
Oh, right. I thought of that before, but forgot again. Thanks for the reminder.
Juergen
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Quoted 9 lines · Jürgen Groß
Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.
Quoted 8 lines
Quoted 2 lines
Could you extend on this? I don't understand what the relevant difference is, from just looking at the changes.
The problem seems to be that a static symbol referenced from the inline asm seems not to silence the error that this static symbol isn't being used. On x86 the
bug_frameis constructed using the %c modifier, which is not supported for Arm (at least gcc 7 used in my compile test complained), but seems to be enough for gcc on x86 to not complain.
But this isn't tied to %c not working on older gcc, is it? It looks instead to be tied to you not specifying the function pointer as an input in the asm(). The compiler would know the symbol is referenced even if the input wasn't used at all in any of the asm()'s operands (but of course it would be better to use the operand once you have it, if it can be used in some way despite %c not being possible to use).
Jan
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Quoted 19 lines · Jan Beulich
Quoted 9 lines · Jürgen Groß
Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.Quoted 8 lines
Quoted 2 lines
Could you extend on this? I don't understand what the relevant difference is, from just looking at the changes.
The problem seems to be that a static symbol referenced from the inline asm seems not to silence the error that this static symbol isn't being used. On x86 the
bug_frameis constructed using the %c modifier, which is not supported for Arm (at least gcc 7 used in my compile test complained), but seems to be enough for gcc on x86 to not complain.
Quoted 7 lines
But this isn't tied to %c not working on older gcc, is it? It looks instead to be tied to you not specifying the function pointer as an input in the
asm(). The compiler would know the symbol is referenced even if the input wasn't used at all in any of the asm()'s operands (but of course it would be better to use the operand once you have it, if it can be used in some way despite %c not being possible to use).
Let me check that.
Juergen
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
In reply to Juergen GrossHi Juergen,
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.
Jan already commented on this, so I am not going to comment again.
Quoted 69 lines
Signed-off-by Juergen Gross <[email protected]>---
V4:
- new patchI have verified the created bugframe is correct by inspecting the created binary.
Signed-off-by Juergen Gross <[email protected]>--- xen/arch/arm/traps.c | 10 +++++++++- xen/drivers/char/ns16550.c | 3 ++- xen/include/asm-arm/bug.h | 32 +++++++++++++++++++++----------- 3 files changed, 32 insertions(+), 13 deletions(-)xen/arch/arm/traps.c +9
-1@@ -1236,8 +1236,16 @@ int do_bug_frame(const struct cpu_user_regs *regs, vaddr_t pc)if ( !bug )return -ENOENT ; +if ( id == BUGFRAME_run_fn ) + { +void (*fn)(const struct cpu_user_regs *) =bug_ptr (bug); + +fn (regs); +return 0 ; + } +/* WARN, BUG or ASSERT: decode the filename pointer and line number. */ - filename =bug_file (bug); + filename =bug_ptr (bug);if ( !is_kernel (filename) )return -EINVAL ; fixup =strlen (filename);xen/drivers/char/ns16550.c +2
-1@@ -192,7 +192,8 @@ static void ns16550_interrupt(/* Safe: ns16550_poll() runs as softirq so not reentrant on a given CPU. */ static DEFINE_PER_CPU(struct serial_port *, poll_port); -static void __ns16550_poll (struct cpu_user_regs *regs) +/* run_in_exception_handler() on Arm requires globally visible symbol. */ +void __ns16550_poll (struct cpu_user_regs *regs) {struct serial_port *port =this_cpu (poll_port);struct ns16550 *uart = port->uart;xen/include/asm-arm/bug.h +2
-2@@ -15,34 +15,38 @@struct bug_frame {signed int loc_disp;/* Relative address to the bug address */ -signed int file_disp;/* Relative address to the filename */ +signed int ptr_disp;/* Relative address to the filename or function */ signed int msg_disp;/* Relative address to the predicate (for ASSERT) */ uint16_t line;/* Line number */ uint32_t pad0:16 ;/* Padding for 8-bytes align */ };#define bug_loc (b) ((const void *)(b) + (b)->loc_disp) -#define bug_file (b) ((const void *)(b) + (b)->file_disp); +#define bug_ptr (b) ((const void *)(b) + (b)->ptr_disp);
xen/include/asm-arm/bug.h line 32
Quoted 10 lines
xen/include/asm-arm/bug.h +4
-3#define bug_line (b) ((b)->line)#define bug_msg (b) ((const char *)(b) + (b)->msg_disp) -#define BUGFRAME_warn 0 -#define BUGFRAME_bug 1 -#define BUGFRAME_assert 2 +#define BUGFRAME_run_fn 0 +#define BUGFRAME_warn 1 +#define BUGFRAME_bug 2 +#define BUGFRAME_assert 3
Why did you renumber it? IOW, why can't BUGFRAME_run_fn be defined as 3?
Quoted 45 lines
xen/include/asm-arm/bug.h +13
-5-#define BUGFRAME_NR 3 +#define BUGFRAME_NR 4/* Many versions of GCC doesn't support the asm %c parameter which would * be preferable to this unpleasantness. We use mergeable string * sections to avoid multiple copies of the string appearing in the * Xen image. */ -#define BUG_FRAME (type, line, file, has_msg, msg) do { \ +#define BUG_FRAME (type, line, ptr, ptr_is_file, has_msg, msg) do { \ BUILD_BUG_ON((line) >> 16); \ BUILD_BUG_ON((type) >= BUGFRAME_NR); \ asm ("1:"BUG_INSTR"\n" \ ".pushsection .rodata.str, \"aMS\", %progbits, 1\n" \ - "2:\t.asciz " __stringify(file) "\n" \ + "2:\n" \ + ".if " #ptr_is_file "\n" \ + "\t.asciz " __stringify(ptr) "\n" \ + ".endif\n" \ "3:\n" \ ".if " #has_msg "\n" \ "\t.asciz " #msg "\n" \@@ -52,21 +56,27 @@ struct bug_frame {"4:\n" \".p2align 2\n" \".long (1b - 4b)\n" \ + ".if " #ptr_is_file "\n" \ ".long (2b - 4b)\n" \ + ".else\n" \ + ".long (" #ptr " - 4b)\n" \ + ".endif\n" \ ".long (3b - 4b)\n" \ ".hword " __stringify(line) ", 0\n" \ ".popsection"); \ } while (0) -#define WARN () BUG_FRAME(BUGFRAME_warn, __LINE__, __FILE__, 0, "") +#define run_in_exception_handler (fn) BUG_FRAME(BUGFRAME_run_fn, 0, fn, 0, 0, "") + +#define WARN () BUG_FRAME(BUGFRAME_warn, __LINE__, __FILE__, 1, 0, "")#define BUG () do { \ - BUG_FRAME(BUGFRAME_bug, __LINE__, __FILE__, 0, ""); \
xen/include/asm-arm/bug.h struct bug_frame { · line 82
Quoted 10 lines
xen/include/asm-arm/bug.h +2
-1+BUG_FRAME (BUGFRAME_bug, __LINE__, __FILE__,1 ,0 ,"" ); \unreachable (); \ } while (0)#define assert_failed (msg) do { \ - BUG_FRAME(BUGFRAME_assert, __LINE__, __FILE__, 1, msg); \ + BUG_FRAME(BUGFRAME_assert, __LINE__, __FILE__, 1, 1, msg); \ unreachable(); \ } while (0)
Cheers,
Signature · 1 line
Julien Grall
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Quoted 11 lines · Julien Grall
Hi Juergen,
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.Jan already commented on this, so I am not going to comment again.
Maybe I can ask some Arm specific question related to this:
In my experiments the only working solution was using the "i" constraint for the function pointer. Do you know whether this is supported for all gcc versions we care about?
Or is there another way to achieve the desired functionality? I'm using now the following macros:
#define BUG_FRAME_run_fn(fn) do { \
asm ("1:"BUG_INSTR"\n" \
".pushsection .bug_frames." __stringify(BUGFRAME_run_fn) \
", \"a\", %%progbits\n" \
"2:\n" \
".p2align 2\n" \
".long (1b - 2b)\n" \
".long (%0 - 2b)\n" \
".long 0\n" \
".hword 0, 0\n" \
".popsection" :: "i" (fn)); \} while (0)
#define run_in_exception_handler(fn) BUG_FRAME_run_fn(fn)
Quoted 72 lines
Quoted 72 lines
Signed-off-by Juergen Gross <[email protected]>---
V4:
- new patchI have verified the created bugframe is correct by inspecting the created binary.
Signed-off-by Juergen Gross <[email protected]>---
xen/arch/arm/traps.c | 10 +++++++++-xen/drivers/char/ns16550.c| 3 ++- xen/include/asm-arm/bug.h | 32 +++++++++++++++++++++-----------3 files changed, 32 insertions(+), 13 deletions(-)xen/arch/arm/traps.c +9
-1@@ -1236,8 +1236,16 @@ int do_bug_frame(const struct cpu_user_regs *regs, vaddr_t pc) if ( !bug ) return -ENOENT; + if ( id == BUGFRAME_run_fn ) + { + void (*fn)(const struct cpu_user_regs *) = bug_ptr(bug); + + fn(regs); + return 0; + } + /* WARN, BUG or ASSERT: decode the filename pointer and line number. */ - filename = bug_file(bug); + filename = bug_ptr(bug); if ( !is_kernel(filename) ) return -EINVAL; fixup = strlen(filename);xen/drivers/char/ns16550.c +1
-1@@ -192,7 +192,8 @@ static void ns16550_interrupt( /* Safe: ns16550_poll() runs as softirq so not reentrant on a given CPU. */ static DEFINE_PER_CPU(struct serial_port *, poll_port); -static void __ns16550_poll (struct cpu_user_regs *regs) +/* run_in_exception_handler() on Arm requires globally visible symbol. */
+void__ns16550_poll(struct cpu_user_regs *regs){
structserial_port*port =this_cpu(poll_port); struct ns16550 *uart =port->uart;xen/include/asm-arm/bug.h +0
-0@@ -15,34 +15,38 @@ struct bug_frame {signed int loc_disp; /* Relative address to the bug address */ - signed int file_disp; /* Relative address to the filename */
+ signed int ptr_disp; /* Relative address to the filename or function */
signed int msg_disp; /* Relative address to the predicate (for ASSERT) */uint16_tline; /* Line number */
uint32_tpad0:16; /* Padding for 8-bytes align */
};
#definebug_loc(b)((const void *)(b) + (b)->loc_disp)xen/include/asm-arm/bug.h +1
-1-#define bug_file (b) ((const void *)(b) + (b)->file_disp); +#define bug_ptr (b) ((const void *)(b) + (b)->ptr_disp);#define
bug_line(b)((b)->line)
#definebug_msg(b)((const char *)(b) + (b)->msg_disp) -#define BUGFRAME_warn 0
Quoted 8 lines
xen/include/asm-arm/bug.h line 32
Quoted 6 lines
xen/include/asm-arm/bug.h +4
-2-#define BUGFRAME_bug 1 -#define BUGFRAME_assert 2 +#define BUGFRAME_run_fn 0 +#define BUGFRAME_warn 1 +#define BUGFRAME_bug 2 +#define BUGFRAME_assert 3Why did you renumber it? IOW, why can't
BUGFRAME_run_fnbe defined as 3?
This matches x86 definition. IMO there is no reason to have a different definition and this will make it more obvious that it might be a good idea to have a common include/xen/bug.h header.
Juergen
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Hi Juergen,
Quoted 11 lines · Jürgen Groß
Quoted 10 lines · Julien Grall
Hi Juergen,
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.
Quoted 7 lines
Quoted 1 line
Jan already commented on this, so I am not going to comment again.
Maybe I can ask some Arm specific question related to this:
In my experiments the only working solution was using the "i" constraint for the function pointer. Do you know whether this is supported for all gcc versions we care about?
I don't know for sure. However, Linux has been using "i" since 2012. So I would assume it ought to be fine for all the version we care.
Quoted 4 lines
Or is there another way to achieve the desired functionality? I'm using now the following macros:
Quoted 12 lines
#define
BUG_FRAME_run_fn(fn)do { \
asm ("1:"BUG_INSTR"\n" \
".pushsection .bug_frames." __stringify(BUGFRAME_run_fn) \
", \"a\", %%progbits\n" \
"2:\n" \
".p2align 2\n" \
".long (1b - 2b)\n" \
".long (%0 - 2b)\n" \
".long 0\n" \
".hword 0, 0\n" \
".popsection" :: "i" (fn)); \
} while (0)
May I ask why we need a new macro?
Quoted 84 lines
#define
run_in_exception_handler(fn)BUG_FRAME_run_fn(fn)Quoted 72 lines
Quoted 72 lines
Signed-off-by Juergen Gross <[email protected]>---
V4:
- new patchI have verified the created bugframe is correct by inspecting the created binary.
Signed-off-by Juergen Gross <[email protected]>---
xen/arch/arm/traps.c | 10 +++++++++-xen/drivers/char/ns16550.c| 3 ++- xen/include/asm-arm/bug.h | 32 +++++++++++++++++++++-----------3 files changed, 32 insertions(+), 13 deletions(-)xen/arch/arm/traps.c +9
-1@@ -1236,8 +1236,16 @@ int do_bug_frame(const struct cpu_user_regs *regs, vaddr_t pc) if ( !bug ) return -ENOENT; + if ( id == BUGFRAME_run_fn ) + { + void (*fn)(const struct cpu_user_regs *) = bug_ptr(bug); + + fn(regs); + return 0; + } + /* WARN, BUG or ASSERT: decode the filename pointer and line number. */ - filename = bug_file(bug); + filename = bug_ptr(bug); if ( !is_kernel(filename) ) return -EINVAL; fixup = strlen(filename);xen/drivers/char/ns16550.c +1
-1@@ -192,7 +192,8 @@ static void ns16550_interrupt( /* Safe: ns16550_poll() runs as softirq so not reentrant on a given CPU. */ static DEFINE_PER_CPU(struct serial_port *, poll_port); -static void __ns16550_poll (struct cpu_user_regs *regs) +/* run_in_exception_handler() on Arm requires globally visible symbol. */
+void__ns16550_poll(struct cpu_user_regs *regs){
structserial_port*port =this_cpu(poll_port); struct ns16550 *uart =port->uart;xen/include/asm-arm/bug.h +0
-0@@ -15,34 +15,38 @@ struct bug_frame {signed int loc_disp; /* Relative address to the bug address */ - signed int file_disp; /* Relative address to the filename */
+ signed int ptr_disp; /* Relative address to the filename or function */
signed int msg_disp; /* Relative address to the predicate (for ASSERT) */uint16_tline; /* Line number */
uint32_tpad0:16; /* Padding for 8-bytes align */
};
#definebug_loc(b)((const void *)(b) + (b)->loc_disp)xen/include/asm-arm/bug.h +1
-1-#define bug_file (b) ((const void *)(b) + (b)->file_disp); +#define bug_ptr (b) ((const void *)(b) + (b)->ptr_disp);#define
bug_line(b)((b)->line)
#definebug_msg(b)((const char *)(b) + (b)->msg_disp) -#define BUGFRAME_warn 0Quoted 8 lines
xen/include/asm-arm/bug.h line 32
Quoted 6 lines
xen/include/asm-arm/bug.h +4
-2-#define BUGFRAME_bug 1 -#define BUGFRAME_assert 2 +#define BUGFRAME_run_fn 0 +#define BUGFRAME_warn 1 +#define BUGFRAME_bug 2 +#define BUGFRAME_assert 3Why did you renumber it? IOW, why can't
BUGFRAME_run_fnbe defined as 3?
Quoted 3 lines
This matches x86 definition. IMO there is no reason to have a different definition and this will make it more obvious that it might be a good idea to have a common
include/xen/bug.hheader.
I agree that common header would be nice. Although, I am not sure if this is achievable. However, my point here is this change would have deserved half-sentence in the commit message because to me this look like unwanted churn.
Cheers,
Signature · 1 line
Julien Grall
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Quoted 41 lines · Julien Grall
Hi Juergen,
Quoted 11 lines · Jürgen Groß
Quoted 10 lines · Julien Grall
Hi Juergen,
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.Quoted 7 lines
Quoted 1 line
Jan already commented on this, so I am not going to comment again.
Maybe I can ask some Arm specific question related to this:
In my experiments the only working solution was using the "i" constraint for the function pointer. Do you know whether this is supported for all gcc versions we care about?
I don't know for sure. However, Linux has been using "i" since 2012. So I would assume it ought to be fine for all the version we care.
Quoted 4 lines
Or is there another way to achieve the desired functionality? I'm using now the following macros:
Quoted 12 lines
#define
BUG_FRAME_run_fn(fn)do { \
asm ("1:"BUG_INSTR"\n" \
".pushsection .bug_frames." __stringify(BUGFRAME_run_fn) \
", \"a\", %%progbits\n" \
"2:\n" \
".p2align 2\n" \
".long (1b - 2b)\n" \
".long (%0 - 2b)\n" \
".long 0\n" \
".hword 0, 0\n" \
".popsection" :: "i" (fn)); \
} while (0)
Quoted 1 line
May I ask why we need a new macro?
Using a common one might be possible, but not with the current way how BUG_FRAME() is defined: gcc complained about the input parameter in case of ASSERT() and WARN().
I might be missing something, but this was the fastest way to at least confirm the scheme is working for Arm.
Quoted 85 lines
Quoted 85 lines
#define
run_in_exception_handler(fn)BUG_FRAME_run_fn(fn)Quoted 80 lines
Quoted 71 lines
Signed-off-by Juergen Gross <[email protected]>---
V4:
- new patchI have verified the created bugframe is correct by inspecting the created binary.
Signed-off-by Juergen Gross <[email protected]>---
xen/arch/arm/traps.c | 10 +++++++++-xen/drivers/char/ns16550.c| 3 ++- xen/include/asm-arm/bug.h | 32 +++++++++++++++++++++-----------3 files changed, 32 insertions(+), 13 deletions(-)xen/arch/arm/traps.c +9
-1@@ -1236,8 +1236,16 @@ int do_bug_frame(const struct cpu_user_regs *regs, vaddr_t pc) if ( !bug ) return -ENOENT; + if ( id == BUGFRAME_run_fn ) + { + void (*fn)(const struct cpu_user_regs *) = bug_ptr(bug); + + fn(regs); + return 0; + } + /* WARN, BUG or ASSERT: decode the filename pointer and line number. */ - filename = bug_file(bug); + filename = bug_ptr(bug); if ( !is_kernel(filename) ) return -EINVAL; fixup = strlen(filename);xen/drivers/char/ns16550.c +1
-1@@ -192,7 +192,8 @@ static void ns16550_interrupt( /* Safe: ns16550_poll() runs as softirq so not reentrant on a given CPU. */ static DEFINE_PER_CPU(struct serial_port *, poll_port); -static void __ns16550_poll (struct cpu_user_regs *regs) +/* run_in_exception_handler() on Arm requires globally visible symbol. */
+void__ns16550_poll(struct cpu_user_regs *regs){
structserial_port*port =this_cpu(poll_port); struct ns16550 *uart =port->uart;xen/include/asm-arm/bug.h +0
-0@@ -15,34 +15,38 @@ struct bug_frame {signed int loc_disp; /* Relative address to the bug address */ - signed int file_disp; /* Relative address to the filename */
+ signed int ptr_disp; /* Relative address to the filename or function */
signed int msg_disp; /* Relative address to the predicate (for ASSERT) */uint16_tline; /* Line number */
uint32_tpad0:16; /* Padding for 8-bytes align */
};
#definebug_loc(b)((const void *)(b) + (b)->loc_disp)xen/include/asm-arm/bug.h +1
-1-#define bug_file (b) ((const void *)(b) + (b)->file_disp); +#define bug_ptr (b) ((const void *)(b) + (b)->ptr_disp);#define
bug_line(b)((b)->line)
#definebug_msg(b)((const char *)(b) + (b)->msg_disp)xen/include/asm-arm/bug.h line 32
Quoted 7 lines
xen/include/asm-arm/bug.h +4
-3-#define BUGFRAME_warn 0 -#define BUGFRAME_bug 1 -#define BUGFRAME_assert 2 +#define BUGFRAME_run_fn 0 +#define BUGFRAME_warn 1 +#define BUGFRAME_bug 2 +#define BUGFRAME_assert 3Why did you renumber it? IOW, why can't
BUGFRAME_run_fnbe defined as 3?
Quoted 5 lines
Quoted 3 lines
This matches x86 definition. IMO there is no reason to have a different definition and this will make it more obvious that it might be a good idea to have a common
include/xen/bug.hheader.I agree that common header would be nice. Although, I am not sure if this is achievable. However, my point here is this change would have deserved half-sentence in the commit message because to me this look like unwanted churn.
Okay.
Juergen
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
Hi Juergen,
Quoted 43 lines · Jürgen Groß
Quoted 42 lines · Julien Grall
Hi Juergen,
Quoted 11 lines · Jürgen Groß
Quoted 10 lines · Julien Grall
Hi Juergen,
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.Quoted 7 lines
Quoted 1 line
Jan already commented on this, so I am not going to comment again.
Maybe I can ask some Arm specific question related to this:
In my experiments the only working solution was using the "i" constraint for the function pointer. Do you know whether this is supported for all gcc versions we care about?
I don't know for sure. However, Linux has been using "i" since 2012. So I would assume it ought to be fine for all the version we care.
Quoted 17 lines
Or is there another way to achieve the desired functionality? I'm using now the following macros:
#define
BUG_FRAME_run_fn(fn)do { \
asm ("1:"BUG_INSTR"\n" \
".pushsection .bug_frames." __stringify(BUGFRAME_run_fn) \
", \"a\", %%progbits\n" \"2:\n" \
".p2align 2\n" \
".long (1b - 2b)\n" \
".long (%0 - 2b)\n" \
".long 0\n" \
".hword 0, 0\n" \
".popsection" :: "i" (fn)); \
} while (0)
Quoted 5 lines
Quoted 1 line
May I ask why we need a new macro?
Using a common one might be possible, but not with the current way how
BUG_FRAME()is defined: gcc complained about the input parameter in case ofASSERT()andWARN().
Could you share the code and the error message?
Quoted 3 lines
I might be missing something, but this was the fastest way to at least confirm the scheme is working for Arm.
Make senses. I also don't have much time to invest in trying to have a common macro. So I am happy with a new macro.
Cheers,
Signature · 1 line
Julien Grall
Re: [PATCH v4 1/3] xen/arm: add support for run_in_exception_handler()
In reply to Jürgen GroßQuoted 43 lines · Jürgen Groß
Quoted 42 lines · Julien Grall
Hi Juergen,
Quoted 11 lines · Jürgen Groß
Quoted 10 lines · Julien Grall
Hi Juergen,
Quoted 6 lines · Juergen Gross
Add support to run a function in an exception handler for Arm. Do it the same way as on x86 via a
bug_frame.Unfortunately inline assembly on Arm seems to be less capable than on x86, leading to functions called via
run_in_exception_handler()having to be globally visible.Quoted 7 lines
Quoted 1 line
Jan already commented on this, so I am not going to comment again.
Maybe I can ask some Arm specific question related to this:
In my experiments the only working solution was using the "i" constraint for the function pointer. Do you know whether this is supported for all gcc versions we care about?
I don't know for sure. However, Linux has been using "i" since 2012. So I would assume it ought to be fine for all the version we care.
Quoted 17 lines
Or is there another way to achieve the desired functionality? I'm using now the following macros:
#define
BUG_FRAME_run_fn(fn)do { \
asm ("1:"BUG_INSTR"\n" \
".pushsection .bug_frames." __stringify(BUGFRAME_run_fn) \
", \"a\", %%progbits\n" \"2:\n" \
".p2align 2\n" \
".long (1b - 2b)\n" \
".long (%0 - 2b)\n" \
".long 0\n" \
".hword 0, 0\n" \
".popsection" :: "i" (fn)); \
} while (0)
Quoted 8 lines
Quoted 1 line
May I ask why we need a new macro?
Using a common one might be possible, but not with the current way how
BUG_FRAME()is defined: gcc complained about the input parameter in case ofASSERT()andWARN().I might be missing something, but this was the fastest way to at least confirm the scheme is working for Arm.
Okay, I think I have found a way to use a common macro, which seems to be even more simple than the original one:
#define BUG_FRAME(type, line, ptr, msg) do { \
BUILD_BUG_ON((line) >> 16); \
BUILD_BUG_ON((type) >= BUGFRAME_NR); \
asm ("1:"BUG_INSTR"\n" \
".pushsection .bug_frames." __stringify(type) \
", \"a\", %%progbits\n" \
"2:\n" \
".p2align 2\n" \
".long (1b - 2b)\n" \
".long (%0 - 2b)\n" \
".long (%1 - 2b)\n" \
".hword " __stringify(line) ", 0\n" \
".popsection" :: "i" (ptr), "i" (msg)); i \} while (0)
Juergen
PATCH v4 2/3 xen: enable keyhandlers to work without register set specified
In reply to Juergen GrossThere are only two keyhandlers which make use of the cpu_user_regs struct passed to them. In order to be able to call any keyhandler in non-interrupt contexts, too, modify those two handlers to copy with a NULL regs pointer by using run_in_exception_handler() in that case.
---
V4:
- new patch
--- xen/common/keyhandler.c | 18 +++++++++++++++--- 1 file changed, 15 insertions(+), 3 deletions(-)
xen/common/keyhandler.c +15 -3
@@ -181,7 +181,10 @@ static void dump_registers(unsigned char key, struct cpu_user_regs *regs)cpumask_copy (&dump_execstate_mask, &cpu_online_map);/* Get local execution state out immediately, in case we get stuck. */ -dump_execstate (regs); +if ( regs ) +dump_execstate (regs); +else +run_in_exception_handler (dump_execstate);/* Alt. handling: remaining CPUs are dumped asynchronously one-by-one. */ if ( alt_key_handling )
@@ -481,15 +484,24 @@ static void run_all_keyhandlers(unsigned char key, struct cpu_user_regs *regs)tasklet_schedule (&run_all_keyhandlers_tasklet); } -static void do_debug_key (unsigned char key,struct cpu_user_regs *regs) +/* run_in_exception_handler() on Arm requires globally visible symbol. */ +void do_debugger_trap_fatal (struct cpu_user_regs *regs) { -printk ("'%c' pressed -> trapping into debugger\n" , key); (void )debugger_trap_fatal (0xf001 , regs);/* Prevent tail call optimisation, which confuses xendbg. */ barrier (); } +static void do_debug_key (unsigned char key,struct cpu_user_regs *regs) +{ +printk ("'%c' pressed -> trapping into debugger\n" , key); +if ( regs ) +do_debugger_trap_fatal (regs); +else +run_in_exception_handler (do_debugger_trap_fatal); +} +static void do_toggle_alt_key (unsigned char key,struct cpu_user_regs *regs) { alt_key_handling = !alt_key_handling;
Signature · 1 line
2.26.2
PATCH v4 3/3 xen: add support for automatic debug key actions in case of crash
In reply to Juergen GrossWhen the host crashes it would sometimes be nice to have additional debug data available which could be produced via debug keys, but halting the server for manual intervention might be impossible due to the need to reboot/kexec rather sooner than later.
Add support for automatic debug key actions in case of crashes which can be activated via boot- or runtime-parameter.
Depending on the type of crash the desired data might be different, so support different settings for the possible types of crashes.
The parameter is "crash-debug" with the following syntax:
crash-debug-<type>=<string>
with <type> being one of:
panic, hwdom, watchdog, kexeccmd, debugkey
and <string> a sequence of debug key characters with '+' having the special semantics of a 10 millisecond pause.
So "crash-debug-watchdog=0+0qr" would result in special output in case of watchdog triggered crash (dom0 state, 10 ms pause, dom0 state, domain info, run queues).
---
V2:
- switched special character '.' to '+' (Jan Beulich)
- 10 ms instead of 1 s pause (Jan Beulich)
- added more text to the boot parameter description (Jan Beulich)
V3:
- added const (Jan Beulich)
- thorough test of crash reason parameter (Jan Beulich)
- kexeccmd case should depend on CONFIG_KEXEC (Jan Beulich)
- added dummy get_irq_regs() helper on Arm
V4:
- call keyhandlers with NULL for regs
- use ARRAY_SIZE() (Jan Beulich)
- don't activate handlers in early boot (Jan Beulich)
- avoid recursion
- extend documentation a bit
--- docs/misc/xen-command-line.pandoc | 41 +++++++++++++++++++++++ xen/common/kexec.c | 8 +++-- xen/common/keyhandler.c | 55 +++++++++++++++++++++++++++++++ xen/common/shutdown.c | 4 +-- xen/drivers/char/console.c | 2 +- xen/include/xen/kexec.h | 10 ++++-- xen/include/xen/keyhandler.h | 10 ++++++ 7 files changed, 122 insertions(+), 8 deletions(-)
docs/misc/xen-command-line.pandoc +41 -0
@@ -574,6 +574,47 @@ reduction of features at Xen's disposal to manage guests. ### cpuinfo (x86) > `= <boolean>` +### crash-debug-debugkey +### crash-debug-hwdom +### crash-debug-kexeccmd +### crash-debug-panic +### crash-debug-watchdog +> `= <string>` + +> Can be modified at runtime + +Specify debug-key actions in cases of crashes. Each of the parameters applies +to a different crash reason. The `<string>` is a sequence of debug key +characters, with `+` having the special meaning of a 10 millisecond pause. + +`crash-debug-debugkey` will be used for crashes induced by the `C` debug +key (i.e. manually induced crash). + +`crash-debug-hwdom` denotes a crash of dom0. + +`crash-debug-kexeccmd` is an explicit request of dom0 to continue with the +kdump kernel via kexec. Only available on hypervisors built with CONFIG_KEXEC. + +`crash-debug-panic` is a crash of the hypervisor. + +`crash-debug-watchdog` is a crash due to the watchdog timer expiring. + +It should be noted that dumping diagnosis data to the console can fail in +multiple ways (missing data, hanging system, ...) depending on the reason +of the crash, which might have left the hypervisor in a bad state. In case +a debug-key action leads to another crash recursion will be avoided, so no +additional debug-key actions will be performed in this case. A crash in the +early boot phase will not result in any debug-key action, as the system +might not yet be in a state where the handlers can work. + +So e.g. `crash-debug-watchdog=0+0r` would dump dom0 state twice with 10 +milliseconds between the two state dumps, followed by the run queues of the +hypervisor, if the system crashes due to a watchdog timeout. + +Depending on the reason of the system crash it might happen that triggering +some debug key action will result in a hang instead of dumping data and then +doing a reboot or crash dump. + ### crashinfo_maxaddr > `= <size>`
xen/common/kexec.c +5 -3
@@ -373,10 +373,12 @@ static int kexec_common_shutdown(void)return 0 ; } -void kexec_crash (void ) +void kexec_crash (enum crash_reason reason) {int pos; +keyhandler_crash_action (reason); + pos = (test_bit (KEXEC_FLAG_CRASH_POS , &kexec_flags) !=0 );if ( !test_bit (KEXEC_IMAGE_CRASH_BASE + pos, &kexec_flags) )return ;
@@ -409,7 +411,7 @@ static long kexec_reboot(void *_image)static void do_crashdump_trigger (unsigned char key) {printk ("'%c' pressed -> triggering crashdump\n" , key); -kexec_crash (); +kexec_crash (CRASHREASON_DEBUGKEY );printk (" * no crash kernel loaded!\n" ); }
@@ -840,7 +842,7 @@ static int kexec_exec(XEN_GUEST_HANDLE_PARAM(void) uarg) ret =continue_hypercall_on_cpu (0 , kexec_reboot, image);break ;case KEXEC_TYPE_CRASH : -kexec_crash ();/* Does not return */ +kexec_crash (CRASHREASON_KEXECCMD );/* Does not return */ break ; }
xen/common/keyhandler.c +55 -0
68 lines not on this page · raw message
xen/common/shutdown.c +2 -2
@@ -43,7 +43,7 @@ void hwdom_shutdown(u8 reason)case SHUTDOWN_crash:debugger_trap_immediate ();printk ("Hardware Dom%u crashed: " , hardware_domain->domain_id); -kexec_crash (); +kexec_crash (CRASHREASON_HWDOM );maybe_reboot ();break ;/* not reached */
@@ -56,7 +56,7 @@ void hwdom_shutdown(u8 reason)case SHUTDOWN_watchdog:printk ("Hardware Dom%u shutdown: watchdog rebooting machine\n" , hardware_domain->domain_id); -kexec_crash (); +kexec_crash (CRASHREASON_WATCHDOG );machine_restart (0 );break ;/* not reached */
xen/drivers/char/console.c +1 -1
@@ -1271,7 +1271,7 @@ void panic(const char *fmt, ...)debugger_trap_immediate (); -kexec_crash (); +kexec_crash (CRASHREASON_PANIC );if ( opt_noreboot )machine_halt ();
xen/include/xen/kexec.h +8 -2
@@ -1,6 +1,8 @@#ifndef __XEN_KEXEC_H__#define __XEN_KEXEC_H__ +#include <xen/keyhandler.h> +#ifdef CONFIG_KEXEC #include <public/kexec.h>
@@ -48,7 +50,7 @@ void machine_kexec_unload(struct kexec_image *image);void machine_kexec_reserved (xen_kexec_reserve_t *reservation);void machine_reboot_kexec (struct kexec_image *image);void machine_kexec (struct kexec_image *image); -void kexec_crash (void ); +void kexec_crash (enum crash_reason reason);void kexec_crash_save_cpu (void );struct crash_xen_info *kexec_crash_save_info (void );void machine_crash_shutdown (void );
@@ -82,7 +84,11 @@ void vmcoreinfo_append_str(const char *fmt, ...)#define kexecing 0static inline void kexec_early_calculations (void ) {} -static inline void kexec_crash (void ) {} +static inline void kexec_crash (enum crash_reason reason) +{ +keyhandler_crash_action (reason); +} +static inline void kexec_crash_save_cpu (void ) {}static inline void set_kexec_crash_area_size (u64 system_ram) {}
xen/include/xen/keyhandler.h +10 -0
@@ -48,4 +48,14 @@ void register_irq_keyhandler(unsigned char key,/* Inject a keypress into the key-handling subsystem. */ extern void handle_keypress (unsigned char key,struct cpu_user_regs *regs); +enum crash_reason { +CRASHREASON_PANIC , +CRASHREASON_HWDOM , +CRASHREASON_WATCHDOG , +CRASHREASON_KEXECCMD , +CRASHREASON_DEBUGKEY , +}; + +void keyhandler_crash_action (enum crash_reason reason); +#endif /* __XEN_KEYHANDLER_H__ */
Signature · 1 line
2.26.2
Quoted 36 lines · Juergen Gross
xen/common/keyhandler.c +32
-0@@ -519,6 +521,59 @@ void __init initialize_keytable(void) } } +#define CRASHACTION_SIZE 32 +static char crash_debug_panic[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-panic" , crash_debug_panic); +static char crash_debug_hwdom[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-hwdom" , crash_debug_hwdom); +static char crash_debug_watchdog[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-watchdog" , crash_debug_watchdog); +#ifdef CONFIG_KEXEC +static char crash_debug_kexeccmd[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-kexeccmd" , crash_debug_kexeccmd); +#else +#define crash_debug_kexeccmd NULL +#endif +static char crash_debug_debugkey[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-debugkey" , crash_debug_debugkey); + +void keyhandler_crash_action (enum crash_reason reason) +{ +static const char *const crash_action[] = { + [CRASHREASON_PANIC ] = crash_debug_panic, + [CRASHREASON_HWDOM ] = crash_debug_hwdom, + [CRASHREASON_WATCHDOG ] = crash_debug_watchdog, + [CRASHREASON_KEXECCMD ] = crash_debug_kexeccmd, + [CRASHREASON_DEBUGKEY ] = crash_debug_debugkey, + }; +static bool ignore; +const char *action; + +/* Some handlers are not functional too early. */ +if ( system_state < SYS_STATE_smp_boot ) +return ; +
xen/common/keyhandler.c void __init initialize_keytable(void) · line 565
Quoted 10 lines
xen/common/keyhandler.c +10
-0+/* Avoid recursion. */ +if ( ignore ) +return ; + ignore = true; + +if ( (unsigned int )reason >=ARRAY_SIZE (crash_action) ) +return ; + action = crash_action[reason]; +if ( !action ) +return ;
If we consider either of the last two "return"s to possibly be taken, I think the "ignore" logic wants to live below them, not above, avoiding no output at all when a recursive invocation turns out to be a "good" one. Then, as before,
Jan
Quoted 52 lines · Jan Beulich
Quoted 36 lines · Juergen Gross
xen/common/keyhandler.c +32
-0@@ -519,6 +521,59 @@ void __init initialize_keytable(void) } } +#define CRASHACTION_SIZE 32 +static char crash_debug_panic[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-panic" , crash_debug_panic); +static char crash_debug_hwdom[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-hwdom" , crash_debug_hwdom); +static char crash_debug_watchdog[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-watchdog" , crash_debug_watchdog); +#ifdef CONFIG_KEXEC +static char crash_debug_kexeccmd[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-kexeccmd" , crash_debug_kexeccmd); +#else +#define crash_debug_kexeccmd NULL +#endif +static char crash_debug_debugkey[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-debugkey" , crash_debug_debugkey); + +void keyhandler_crash_action (enum crash_reason reason) +{ +static const char *const crash_action[] = { + [CRASHREASON_PANIC ] = crash_debug_panic, + [CRASHREASON_HWDOM ] = crash_debug_hwdom, + [CRASHREASON_WATCHDOG ] = crash_debug_watchdog, + [CRASHREASON_KEXECCMD ] = crash_debug_kexeccmd, + [CRASHREASON_DEBUGKEY ] = crash_debug_debugkey, + }; +static bool ignore; +const char *action; + +/* Some handlers are not functional too early. */ +if ( system_state < SYS_STATE_smp_boot ) +return ; +xen/common/keyhandler.c void __init initialize_keytable(void) · line 565
Quoted 10 lines
xen/common/keyhandler.c +10
-0+/* Avoid recursion. */ +if ( ignore ) +return ; + ignore = true; + +if ( (unsigned int )reason >=ARRAY_SIZE (crash_action) ) +return ; + action = crash_action[reason]; +if ( !action ) +return ;If we consider either of the last two "return"s to possibly be taken, I think the "ignore" logic wants to live below them, not above, avoiding no output at all when a recursive invocation turns out to be a "good" one. Then, as before,
Quoted 1 line
Reviewed-by Jan Beulich <[email protected]>
Fine with me.
Juergen
Hi Juergen,
Quoted 37 lines · Juergen Gross
xen/common/keyhandler.c +21
-0@@ -3,7 +3,9 @@*/ #include <asm/regs.h> +#include <xen/delay.h> #include <xen/keyhandler.h> +#include <xen/param.h> #include <xen/shutdown.h> #include <xen/event.h> #include <xen/console.h> @@ -519,6 +521,59 @@ void __init initialize_keytable(void) } } +#define CRASHACTION_SIZE 32 +static char crash_debug_panic[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-panic" , crash_debug_panic); +static char crash_debug_hwdom[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-hwdom" , crash_debug_hwdom); +static char crash_debug_watchdog[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-watchdog" , crash_debug_watchdog); +#ifdef CONFIG_KEXEC +static char crash_debug_kexeccmd[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-kexeccmd" , crash_debug_kexeccmd); +#else +#define crash_debug_kexeccmd NULL +#endif +static char crash_debug_debugkey[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-debugkey" , crash_debug_debugkey); + +void keyhandler_crash_action (enum crash_reason reason) +{ +static const char *const crash_action[] = {
xen/common/keyhandler.c void __init initialize_keytable(void) · line 552
Quoted 10 lines
xen/common/keyhandler.c +10
-0+ [CRASHREASON_PANIC] = crash_debug_panic, + [CRASHREASON_HWDOM] = crash_debug_hwdom, + [CRASHREASON_WATCHDOG] = crash_debug_watchdog, + [CRASHREASON_KEXECCMD] = crash_debug_kexeccmd, + [CRASHREASON_DEBUGKEY] = crash_debug_debugkey, + }; +static bool ignore; +const char *action; + +/* Some handlers are not functional too early. */
Can you explain in the commit message why this is necessary (An example would be useful)?
Cheers,
Signature · 1 line
Julien Grall
Quoted 42 lines · Julien Grall
Hi Juergen,
Quoted 39 lines · Juergen Gross
xen/common/keyhandler.c +25
-0@@ -3,7 +3,9 @@*/ #include <asm/regs.h> +#include <xen/delay.h> #include <xen/keyhandler.h> +#include <xen/param.h> #include <xen/shutdown.h> #include <xen/event.h> #include <xen/console.h>@@ -519,6 +521,59 @@ void __init initialize_keytable(void) } } +#define CRASHACTION_SIZE 32 +static char crash_debug_panic[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-panic" , crash_debug_panic); +static char crash_debug_hwdom[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-hwdom" , crash_debug_hwdom); +static char crash_debug_watchdog[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-watchdog" , crash_debug_watchdog); +#ifdef CONFIG_KEXEC +static char crash_debug_kexeccmd[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-kexeccmd" , crash_debug_kexeccmd); +#else +#define crash_debug_kexeccmd NULL +#endif +static char crash_debug_debugkey[CRASHACTION_SIZE ]; +string_runtime_param ("crash-debug-debugkey" , crash_debug_debugkey); + +void keyhandler_crash_action (enum crash_reason reason) +{ + static const char *const crash_action[] = { + [CRASHREASON_PANIC] = crash_debug_panic, + [CRASHREASON_HWDOM] = crash_debug_hwdom, + [CRASHREASON_WATCHDOG] = crash_debug_watchdog, + [CRASHREASON_KEXECCMD] = crash_debug_kexeccmd,
Quoted 8 lines
xen/common/keyhandler.c void __init initialize_keytable(void) · line 552
Quoted 6 lines
xen/common/keyhandler.c +6
-0+ [CRASHREASON_DEBUGKEY] = crash_debug_debugkey, + }; + static bool ignore; + const char *action; + + /* Some handlers are not functional too early. */Can you explain in the commit message why this is necessary (An example would be useful)?
Okay.
Juergen
Quoted 5 lines · Juergen Gross
Patch 2 opens up more potential for simplification: in theory there is no need any more to call any key handler with the regs parameter, allowing to use the same prototype for all handlers. The downside would be to have an additional irq frame on the stack for the
dump_registers()and thedo_debug_key()handlers.
This isn't the only downside, is it? We'd then also need to be able to (sufficiently cleanly) unwind through the new frame to reach the prior one, in order to avoid logging less reliable information. Plus decompose the prior frame as well to avoid logging less easy to consume data.
Jan
Quoted 7 lines · Jan Beulich
Quoted 5 lines · Juergen Gross
Patch 2 opens up more potential for simplification: in theory there is no need any more to call any key handler with the regs parameter, allowing to use the same prototype for all handlers. The downside would be to have an additional irq frame on the stack for the
dump_registers()and thedo_debug_key()handlers.
Quoted 5 lines
This isn't the only downside, is it? We'd then also need to be able to (sufficiently cleanly) unwind through the new frame to reach the prior one, in order to avoid logging less reliable information. Plus decompose the prior frame as well to avoid logging less easy to consume data.
Yes, this was implied by the "additional irq frame on the stack".
Juergen
Quoted 8 lines · Jürgen Groß
Quoted 7 lines · Jan Beulich
Quoted 5 lines · Juergen Gross
Patch 2 opens up more potential for simplification: in theory there is no need any more to call any key handler with the regs parameter, allowing to use the same prototype for all handlers. The downside would be to have an additional irq frame on the stack for the
dump_registers()and thedo_debug_key()handlers.
Quoted 7 lines
Quoted 5 lines
This isn't the only downside, is it? We'd then also need to be able to (sufficiently cleanly) unwind through the new frame to reach the prior one, in order to avoid logging less reliable information. Plus decompose the prior frame as well to avoid logging less easy to consume data.
Yes, this was implied by the "additional irq frame on the stack".
Oh, okay - I read it as just referring to the possible concern of more of the not overly large stack to get used.
Jan