12:31 a.m.
Hi Lianbo,
I thought of an idea that makes this patch better, especially support
transitional versions too, so please hold on.
Thanks,
Kazu
On 2023/05/16 16:25, HAGIO KAZUHITO(萩尾 一仁) wrote:
> From: Kazuhito Hagio <k-hagio-ab(a)nec.com>
>
> Kernel commit fb799447ae29 ("x86,objtool: Split UNWIND_HINT_EMPTY in
> two"), which is contained in Linux 6.4 and later kernels, changed
> ORC_TYPE_CALL macro from 0 to 2. As a result, the "bt" command cannot
> use ORC entries and displays stale entries in a call trace.
>
> crash> bt 1
> PID: 1 TASK: ffff93cd06294180 CPU: 51 COMMAND: "systemd"
> #0 [ffffb72bc00cbc98] __schedule at ffffffff86e52aae
> #1 [ffffb72bc00cbd00] schedule at ffffffff86e52f6a
> #2 [ffffb72bc00cbd18] schedule_hrtimeout_range_clock at ffffffff86e58ef5
> #3 [ffffb72bc00cbd88] ep_poll at ffffffff8669624d
> #4 [ffffb72bc00cbe28] do_epoll_wait at ffffffff86696371
> #5 [ffffb72bc00cbe30] do_timerfd_settime at ffffffff8669902b <<
> #6 [ffffb72bc00cbe60] __x64_sys_epoll_wait at ffffffff86696bf0
> #7 [ffffb72bc00cbeb0] do_syscall_64 at ffffffff86e3feb9
> #8 [ffffb72bc00cbee0] __task_pid_nr_ns at ffffffff863330d7 <<
> #9 [ffffb72bc00cbf08] syscall_exit_to_user_mode at ffffffff86e466b2 <<
stale entries
> #10 [ffffb72bc00cbf18] do_syscall_64 at ffffffff86e3fec9 <<
> #11 [ffffb72bc00cbf50] entry_SYSCALL_64_after_hwframe at ffffffff870000aa
>
> Also, struct orc_entry in kernel has changed, and debugging information
> for ORC unwinder can be displayed incorrectly.
>
> To fix these,
> (1) introduce "kernel_orc_entry_6_4" structure corresponding to 6.4 and
> abstruct layer "orc_entry" structure in crash,
> (2) switch ORC_TYPE_CALL to 2 or 0 with kernel's orc_entry structure.
>
> Signed-off-by: Kazuhito Hagio <k-hagio-ab(a)nec.com>
> ---
> defs.h | 24 +++++++++++--
> x86_64.c | 104 ++++++++++++++++++++++++++++++++++++++++++-------------
> 2 files changed, 102 insertions(+), 26 deletions(-)
>
> diff --git a/defs.h b/defs.h
> index 211fc9d55d33..0cdfa295f964 100644
> --- a/defs.h
> +++ b/defs.h
> @@ -6356,6 +6356,25 @@ typedef struct __attribute__((__packed__)) {
> unsigned int end:1;
> } kernel_orc_entry;
>
> +typedef struct __attribute__((__packed__)) {
> + signed short sp_offset;
> + signed short bp_offset;
> + unsigned int sp_reg:4;
> + unsigned int bp_reg:4;
> + unsigned int type:3;
> + unsigned int signal:1;
> +} kernel_orc_entry_6_4;
> +
> +typedef struct orc_entry {
> + signed short sp_offset;
> + signed short bp_offset;
> + unsigned int sp_reg;
> + unsigned int bp_reg;
> + unsigned int type;
> + unsigned int end;
> + unsigned int signal;
> +} orc_entry;
> +
> struct ORC_data {
> int module_ORC;
> uint lookup_num_blocks;
> @@ -6366,10 +6385,10 @@ struct ORC_data {
> ulong orc_lookup;
> ulong ip_entry;
> ulong orc_entry;
> - kernel_orc_entry kernel_orc_entry;
> + orc_entry orc_entry_data;
> };
>
> -#define ORC_TYPE_CALL 0
> +#define ORC_TYPE_CALL ((machdep->flags & ORC_6_4) ? 2 :
0)
> #define ORC_TYPE_REGS 1
> #define ORC_TYPE_REGS_IRET 2
> #define UNWIND_HINT_TYPE_SAVE 3
> @@ -6446,6 +6465,7 @@ struct machine_specific {
> #define ORC (0x4000)
> #define KPTI (0x8000)
> #define L1TF (0x10000)
> +#define ORC_6_4 (0x20000)
>
> #define VM_FLAGS (VM_ORIG|VM_2_6_11|VM_XEN|VM_XEN_RHEL4|VM_5LEVEL)
>
> diff --git a/x86_64.c b/x86_64.c
> index 5019c69e452e..7d863848e178 100644
> --- a/x86_64.c
> +++ b/x86_64.c
> @@ -132,9 +132,9 @@ static void GART_init(void);
> static void x86_64_exception_stacks_init(void);
> static int in_START_KERNEL_map(ulong);
> static ulong orc_ip(ulong);
> -static kernel_orc_entry *__orc_find(ulong, ulong, uint, ulong);
> -static kernel_orc_entry *orc_find(ulong);
> -static kernel_orc_entry *orc_module_find(ulong);
> +static orc_entry *__orc_find(ulong, ulong, uint, ulong);
> +static orc_entry *orc_find(ulong);
> +static orc_entry *orc_module_find(ulong);
> static ulong ip_table_to_vaddr(ulong);
> static void orc_dump(ulong);
>
> @@ -806,6 +806,8 @@ x86_64_dump_machdep_table(ulong arg)
> fprintf(fp, "%sFRAMESIZE_DEBUG", others++ ? "|" :
"");
> if (machdep->flags & ORC)
> fprintf(fp, "%sORC", others++ ? "|" : "");
> + if (machdep->flags & ORC_6_4)
> + fprintf(fp, "%sORC_6_4", others++ ? "|" : "");
> if (machdep->flags & FRAMEPOINTER)
> fprintf(fp, "%sFRAMEPOINTER", others++ ? "|" :
"");
> if (machdep->flags & GART_REGION)
> @@ -988,16 +990,20 @@ x86_64_dump_machdep_table(ulong arg)
> fprintf(fp, " orc_lookup: %lx\n",
ms->orc.orc_lookup);
> fprintf(fp, " ip_entry: %lx\n",
ms->orc.ip_entry);
> fprintf(fp, " orc_entry: %lx\n",
ms->orc.orc_entry);
> - fprintf(fp, " kernel_orc_entry:\n");
> - fprintf(fp, " sp_offset: %d\n",
ms->orc.kernel_orc_entry.sp_offset);
> - fprintf(fp, " bp_offset: %d\n",
ms->orc.kernel_orc_entry.bp_offset);
> - fprintf(fp, " sp_reg: %d\n",
ms->orc.kernel_orc_entry.sp_reg);
> - fprintf(fp, " bp_reg: %d\n",
ms->orc.kernel_orc_entry.bp_reg);
> - fprintf(fp, " type: %d\n",
ms->orc.kernel_orc_entry.type);
> + fprintf(fp, " orc_entry_data:\n");
> + fprintf(fp, " sp_offset: %d\n",
ms->orc.orc_entry_data.sp_offset);
> + fprintf(fp, " bp_offset: %d\n",
ms->orc.orc_entry_data.bp_offset);
> + fprintf(fp, " sp_reg: %d\n",
ms->orc.orc_entry_data.sp_reg);
> + fprintf(fp, " bp_reg: %d\n",
ms->orc.orc_entry_data.bp_reg);
> + fprintf(fp, " type: %d\n",
ms->orc.orc_entry_data.type);
> if (MEMBER_EXISTS("orc_entry", "end"))
> - fprintf(fp, " end: %d\n",
ms->orc.kernel_orc_entry.end);
> + fprintf(fp, " end: %d\n",
ms->orc.orc_entry_data.end);
> else
> fprintf(fp, " end: (n/a)\n");
> + if (MEMBER_EXISTS("orc_entry", "signal"))
> + fprintf(fp, " signal: %d\n",
ms->orc.orc_entry_data.signal);
> + else
> + fprintf(fp, " signal: (n/a)\n");
> }
> fprintf(fp, " pto: %s",
> machdep->flags & PT_REGS_INIT ? "\n" :
"(uninitialized)\n");
> @@ -6391,7 +6397,8 @@ x86_64_ORC_init(void)
> !MEMBER_EXISTS("orc_entry", "sp_reg") ||
> !MEMBER_EXISTS("orc_entry", "bp_reg") ||
> !MEMBER_EXISTS("orc_entry", "type") ||
> - SIZE(orc_entry) != sizeof(kernel_orc_entry)) {
> + (SIZE(orc_entry) != sizeof(kernel_orc_entry) &&
> + SIZE(orc_entry) != sizeof(kernel_orc_entry_6_4))) {
> error(WARNING, "ORC unwinder: orc_entry structure has changed\n");
> return;
> }
> @@ -6439,6 +6446,10 @@ x86_64_ORC_init(void)
> MEMBER_OFFSET_INIT(inactive_task_frame_bp, "inactive_task_frame",
"bp");
> MEMBER_OFFSET_INIT(inactive_task_frame_ret_addr, "inactive_task_frame",
"ret_addr");
>
> + if (MEMBER_EXISTS("orc_entry", "signal") && /* added at
6.3 */
> + !MEMBER_EXISTS("orc_entry", "end")) /* removed at 6.4
with type change */
> + machdep->flags |= ORC_6_4;
> +
> machdep->flags |= ORC;
> }
>
> @@ -8521,7 +8532,7 @@ x86_64_get_framesize(struct bt_info *bt, ulong textaddr, ulong
rsp, char *stack_
> int reterror;
> int arg_exists;
> int exception;
> - kernel_orc_entry *korc;
> + orc_entry *korc;
>
> if (!(bt->flags & BT_FRAMESIZE_DEBUG)) {
> if ((bt->flags & BT_FRAMESIZE_IGNORE_MASK) ||
> @@ -8613,6 +8624,8 @@ x86_64_get_framesize(struct bt_info *bt, ulong textaddr, ulong
rsp, char *stack_
> korc->sp_reg, korc->bp_reg, korc->type);
> if (MEMBER_EXISTS("orc_entry", "end"))
> fprintf(fp, " end: %d", korc->end);
> + if (MEMBER_EXISTS("orc_entry", "signal"))
> + fprintf(fp, " signal: %d", korc->signal);
> fprintf(fp, "\n");
> }
>
> @@ -9117,7 +9130,43 @@ orc_ip(ulong ip)
> return (ip + ip_entry);
> }
>
> -static kernel_orc_entry *
> +static orc_entry *
> +orc_get_entry(struct ORC_data *orc)
> +{
> + struct orc_entry *entry = &orc->orc_entry_data;
> +
> + if (machdep->flags & ORC_6_4) {
> + kernel_orc_entry_6_4 korc;
> +
> + if (!readmem(orc->orc_entry, KVADDR, &korc, sizeof(kernel_orc_entry_6_4),
> + "kernel orc_entry", RETURN_ON_ERROR|QUIET))
> + return NULL;
> +
> + entry->sp_offset = korc.sp_offset;
> + entry->bp_offset = korc.bp_offset;
> + entry->sp_reg = korc.sp_reg;
> + entry->bp_reg = korc.bp_reg;
> + entry->type = korc.type;
> + entry->signal = korc.signal;
> + } else {
> + kernel_orc_entry korc;
> +
> + if (!readmem(orc->orc_entry, KVADDR, &korc, sizeof(kernel_orc_entry),
> + "kernel orc_entry", RETURN_ON_ERROR|QUIET))
> + return NULL;
> +
> + entry->sp_offset = korc.sp_offset;
> + entry->bp_offset = korc.bp_offset;
> + entry->sp_reg = korc.sp_reg;
> + entry->bp_reg = korc.bp_reg;
> + entry->type = korc.type;
> + entry->end = korc.end;
> + }
> +
> + return entry;
> +}
> +
> +static orc_entry *
> __orc_find(ulong ip_table_ptr, ulong u_table_ptr, uint num_entries, ulong ip)
> {
> int index;
> @@ -9127,7 +9176,7 @@ __orc_find(ulong ip_table_ptr, ulong u_table_ptr, uint
num_entries, ulong ip)
> int *ip_table = (int *)ip_table_ptr;
> struct ORC_data *orc = &machdep->machspec->orc;
> ulong vaddr;
> - kernel_orc_entry *korc;
> + orc_entry *korc;
>
> if (CRASHDEBUG(2)) {
> int i, ip_entry;
> @@ -9171,11 +9220,11 @@ __orc_find(ulong ip_table_ptr, ulong u_table_ptr, uint
num_entries, ulong ip)
>
> orc->ip_entry = (ulong)found;
> orc->orc_entry = u_table_ptr + (index * SIZE(orc_entry));
> - if (!readmem(orc->orc_entry, KVADDR, &orc->kernel_orc_entry,
> - sizeof(kernel_orc_entry), "kernel orc_entry",
RETURN_ON_ERROR|QUIET))
> +
> + if (!orc_get_entry(orc))
> return NULL;
>
> - korc = &orc->kernel_orc_entry;
> + korc = &orc->orc_entry_data;
>
> if (CRASHDEBUG(2)) {
> fprintf(fp, " found: %lx index: %d\n", (ulong)found, index);
> @@ -9184,6 +9233,8 @@ __orc_find(ulong ip_table_ptr, ulong u_table_ptr, uint
num_entries, ulong ip)
> orc->orc_entry, korc->sp_offset, korc->bp_offset, korc->sp_reg,
korc->bp_reg, korc->type);
> if (MEMBER_EXISTS("orc_entry", "end"))
> fprintf(fp, " end: %d", korc->end);
> + if (MEMBER_EXISTS("orc_entry", "signal"))
> + fprintf(fp, " signal: %d", korc->signal);
> fprintf(fp, "\n");
> }
>
> @@ -9195,7 +9246,7 @@ __orc_find(ulong ip_table_ptr, ulong u_table_ptr, uint
num_entries, ulong ip)
> #define LOOKUP_START_IP (unsigned long)kt->stext
> #define LOOKUP_STOP_IP (unsigned long)kt->etext
>
> -static kernel_orc_entry *
> +static orc_entry *
> orc_find(ulong ip)
> {
> unsigned int idx, start, stop;
> @@ -9265,7 +9316,7 @@ orc_find(ulong ip)
> orc->__start_orc_unwind + (start * SIZE(orc_entry)), stop - start, ip);
> }
>
> -static kernel_orc_entry *
> +static orc_entry *
> orc_module_find(ulong ip)
> {
> struct load_module *lm;
> @@ -9312,7 +9363,7 @@ static void
> orc_dump(ulong ip)
> {
> struct ORC_data *orc = &machdep->machspec->orc;
> - kernel_orc_entry *korc;
> + orc_entry *korc;
> ulong vaddr, offset;
> struct syment *sp, *orig;
>
> @@ -9335,18 +9386,23 @@ next_in_func:
> fprintf(fp, "%s+%ld -> ", sp->name, offset);
> else
> fprintf(fp, "(unresolved) -> ");
> - if (!readmem(orc->orc_entry, KVADDR, &orc->kernel_orc_entry,
sizeof(kernel_orc_entry),
> - "kernel orc_entry", RETURN_ON_ERROR))
> +
> + if (!orc_get_entry(orc))
> error(FATAL, "cannot read orc_entry\n");
> - korc = &orc->kernel_orc_entry;
> + korc = &orc->orc_entry_data;
> fprintf(fp, "orc: %lx spo: %d bpo: %d spr: %d bpr: %d type: %d",
> orc->orc_entry, korc->sp_offset, korc->bp_offset, korc->sp_reg,
korc->bp_reg, korc->type);
> if (MEMBER_EXISTS("orc_entry", "end"))
> fprintf(fp, " end: %d", korc->end);
> + if (MEMBER_EXISTS("orc_entry", "signal"))
> + fprintf(fp, " signal: %d", korc->signal);
> fprintf(fp, "\n");
>
> orc->ip_entry += sizeof(int);
> - orc->orc_entry += sizeof(kernel_orc_entry);
> + if (machdep->flags & ORC_6_4)
> + orc->orc_entry += sizeof(kernel_orc_entry_6_4);
> + else
> + orc->orc_entry += sizeof(kernel_orc_entry);
> vaddr = ip_table_to_vaddr(orc->ip_entry);
> if ((sp = value_search(vaddr, &offset)))
> if (sp == orig)