kdb: Don't implicitly change tasks; plus misc fixups
Subsystems (2)
Douglas Anderson
reviewercover kdb: Don't implicitly change tasks; plus misc fixups
This started out as just a follow-up to Daniel's post [1]. I wanted to stop implicitly changing the current task in kdb. ...but, of course, everywhere you look in kdb there are things to cleanup, so this gets a few misc cleanups I found along the way. Enjoy.
[1] https://lore.kernel.org/r/[email protected]
Douglas Anderson (5):
MIPS: kdb: Remove old workaround for backtracing on other CPUs
kdb: kdb_current_regs should be private
kdb: kdb_current_task shouldn't be exported
kdb: Gid rid of implicit setting of the current task / regs
kdb: Get rid of confusing diag msg from "rd" if current task has no
regs
arch/mips/kernel/traps.c | 5 ----- include/linux/kdb.h | 2 -- kernel/debug/kdb/kdb_bt.c | 8 +------- kernel/debug/kdb/kdb_main.c | 31 ++++++++++++++----------------- kernel/debug/kdb/kdb_private.h | 2 +- 5 files changed, 16 insertions(+), 32 deletions(-)
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
Douglas Anderson
reviewerPATCH 1/5 MIPS: kdb: Remove old workaround for backtracing on other CPUs
As of commit 2277b492582d ("kdb: Fix stack crawling on 'running' CPUs that aren't the master") we no longer need any special case for doing stack dumps on CPUs that are not the kdb master. Let's remove.
---
I have no way to test this personally, so hopefully someone who uses kdb/kgdb on MIPS can.
Ideally this patch should be Acked by MIPS folks and then land through the kdb/kgdb tree since the next patch in the series, ("kdb:kdb_current_regs should be private") depends on it.
arch/mips/kernel/traps.c | 5 ----- 1 file changed, 5 deletions(-)
arch/mips/kernel/traps.c +0 -5
@@ -210,11 +210,6 @@ void show_stack(struct task_struct *task, unsigned long *sp) regs.regs[29 ] = task->thread.reg29; regs.regs[31 ] =0 ; regs.cp0_epc = task->thread.reg31; -#ifdef CONFIG_KGDB_KDB - }else if (atomic_read (&kgdb_active) !=-1 && - kdb_current_regs) { -memcpy (®s, kdb_current_regs,sizeof (regs)); -#endif /* CONFIG_KGDB_KDB */ }else {prepare_frametrace (®s); }
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
Re: [PATCH 1/5] MIPS: kdb: Remove old workaround for backtracing on other CPUs
Quoted 8 lines · Douglas Anderson
As of commit
2277b492582d("kdb: Fix stack crawling on 'running' CPUs that aren't the master") we no longer need any special case for doing stack dumps on CPUs that are not the kdb master. Let's remove.Signed-off-by Douglas Anderson <[email protected]>---
I have no way to test this personally, so hopefully someone who uses kdb/kgdb on MIPS can.
I took this as a hint to add mips support to kgdbtest ;-)
Support is added and working well. Unfortunately lack of familiarity with mips means I have not yet figured out which mips defconfig gives us working SMP (and what the corresponding qemu invocation should be).
I think that means I still can't (quite) exercise this code fully. The most appropriate test is bta on an SMP system, right?
Quoted 3 lines
Ideally this patch should be Acked by MIPS folks and then land through the kdb/kgdb tree since the next patch in the series, ("kdb:
kdb_current_regsshould be private") depends on it.
An Acked-by from a MIPS maintainer would be very welcome. Perhaps with a bit of extra work on the above I might be able to provide a Tested-by:.
I didn't see anything that particularly bothered me in the patches but given we're already at -rc7 I'm inclined to target this patchset for 5.6 rather than 5.5.
Daniel.
Doug Anderson
reviewerRe: [PATCH 1/5] MIPS: kdb: Remove old workaround for backtracing on other CPUs
Hi,
Quoted 11 lines · Daniel Thompson
Quoted 8 lines · Douglas Anderson
As of commit
2277b492582d("kdb: Fix stack crawling on 'running' CPUs that aren't the master") we no longer need any special case for doing stack dumps on CPUs that are not the kdb master. Let's remove.Signed-off-by Douglas Anderson <[email protected]>---
I have no way to test this personally, so hopefully someone who uses kdb/kgdb on MIPS can.
Quoted 5 lines
I took this as a hint to add mips support to kgdbtest ;-)
Support is added and working well. Unfortunately lack of familiarity with mips means I have not yet figured out which mips defconfig gives us working SMP (and what the corresponding qemu invocation should be).
Nice!
Quoted 2 lines
I think that means I still can't (quite) exercise this code fully. The most appropriate test is bta on an SMP system, right?
Yeah, or at least "btc".
Quoted 11 lines
Quoted 3 lines
Ideally this patch should be Acked by MIPS folks and then land through the kdb/kgdb tree since the next patch in the series, ("kdb:
kdb_current_regsshould be private") depends on it.An Acked-by from a MIPS maintainer would be very welcome. Perhaps with a bit of extra work on the above I might be able to provide a Tested-by:.
I didn't see anything that particularly bothered me in the patches but given we're already at -rc7 I'm inclined to target this patchset for 5.6 rather than 5.5.
That's fine. This is all just cleanup stuff so targeting 5.6 is fine.
-Doug
Re: [PATCH 1/5] MIPS: kdb: Remove old workaround for backtracing on other CPUs
In reply to Daniel ThompsonHi Daniel,
Quoted 11 lines · Daniel Thompson
Quoted 8 lines · Douglas Anderson
As of commit
2277b492582d("kdb: Fix stack crawling on 'running' CPUs that aren't the master") we no longer need any special case for doing stack dumps on CPUs that are not the kdb master. Let's remove.Signed-off-by Douglas Anderson <[email protected]>---
I have no way to test this personally, so hopefully someone who uses kdb/kgdb on MIPS can.I took this as a hint to add mips support to kgdbtest ;-)
Wonderful! :)
Quoted 3 lines
Support is added and working well. Unfortunately lack of familiarity with mips means I have not yet figured out which mips defconfig gives us working SMP (and what the corresponding qemu invocation should be).
You can build 64r6el_defconfig & boot it something like this:
$ qemu-system-mips64el \
-M boston -cpu I6500 -smp 4 \
-kernel arch/mips/boot/vmlinux.gz.itb \
-serial stdio \
-hda my-disk-image.bin \
-append "root=/dev/sda"Linux should see the system as a single core with 4 hardware threads (VPs or Virtual Processors in MIPS terminology).
Quoted 7 lines
Quoted 3 lines
Ideally this patch should be Acked by MIPS folks and then land through the kdb/kgdb tree since the next patch in the series, ("kdb:
kdb_current_regsshould be private") depends on it.An Acked-by from a MIPS maintainer would be very welcome. Perhaps with a bit of extra work on the above I might be able to provide a Tested-by:.
The patches look reasonable to me; I was hoping to test them before giving an ack but haven't had the time yet. It seems you may be making that easier :)
Thanks,
Paul
Douglas Anderson
reviewerPATCH 2/5 kdb: kdb_current_regs should be private
In reply to Douglas AndersonAs of the patch ("MIPS: kdb: Remove old workaround for backtracing on other CPUs") there is no reason for kdb_current_regs to be in the public "kdb.h". Let's move it next to kdb_current_task.
---
include/linux/kdb.h | 2 -- kernel/debug/kdb/kdb_private.h | 1 + 2 files changed, 1 insertion(+), 2 deletions(-)
include/linux/kdb.h +0 -2
@@ -183,8 +183,6 @@ int kdb_process_cpu(const struct task_struct *p)return cpu; } -/* kdb access to register set for stack dumping */ -extern struct pt_regs *kdb_current_regs;#ifdef CONFIG_KALLSYMS extern const char *kdb_walk_kallsyms (loff_t *pos);#else /* ! CONFIG_KALLSYMS */
kernel/debug/kdb/kdb_private.h +1 -0
@@ -242,6 +242,7 @@ extern void debug_kusage(void);extern void kdb_set_current_task (struct task_struct *);extern struct task_struct *kdb_current_task; +extern struct pt_regs *kdb_current_regs;#ifdef CONFIG_KDB_KEYBOARD extern void kdb_kbd_cleanup_state (void );
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
Douglas Anderson
reviewerPATCH 3/5 kdb: kdb_current_task shouldn't be exported
In reply to Douglas AndersonThe kdb_current_task variable has been declared in "kernel/debug/kdb/kdb_private.h" since 2010 when kdb was added to the mainline kernel. This is not a public header. There should be no reason that kdb_current_task should be exported and there are no in-kernel users that need it. Remove the export.
---
kernel/debug/kdb/kdb_main.c | 1 - 1 file changed, 1 deletion(-)
kernel/debug/kdb/kdb_main.c +0 -1
@@ -73,7 +73,6 @@ int kdb_nextline = 1;int kdb_state;/* General KDB state */ struct task_struct *kdb_current_task; -EXPORT_SYMBOL (kdb_current_task);struct pt_regs *kdb_current_regs;const char *kdb_diemsg;
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
Douglas Anderson
reviewerPATCH 4/5 kdb: Gid rid of implicit setting of the current task / regs
In reply to Douglas AndersonSome (but not all?) of the kdb backtrace paths would cause the kdb_current_task and kdb_current_regs to remain changed. As discussed in a review of a previous patch [1], this doesn't seem intuitive, so let's fix that.
...but, it turns out that there's actually no longer any reason to set the current task / current regs while backtracing anymore anyway. As of commit 2277b492582d ("kdb: Fix stack crawling on 'running' CPUs that aren't the master") if we're backtracing on a task running on a CPU we ask that CPU to do the backtrace itself. Linux can do that without anything fancy. If we're doing backtrace on a sleeping task we can also do that fine without updating globals. So this patch mostly just turns into deleting a bunch of code.
[1] https://lore.kernel.org/r/[email protected]
---
kernel/debug/kdb/kdb_bt.c | 8 +------- kernel/debug/kdb/kdb_main.c | 2 +- kernel/debug/kdb/kdb_private.h | 1 - 3 files changed, 2 insertions(+), 9 deletions(-)
kernel/debug/kdb/kdb_bt.c +1 -7
@@ -119,7 +119,6 @@ kdb_bt_cpu(unsigned long cpu)return ; } -kdb_set_current_task (kdb_tsk);kdb_bt1 (kdb_tsk, ~0UL , false); }
@@ -166,10 +165,8 @@ kdb_bt(int argc, const char **argv)if (diag)return diag; p =find_task_by_pid_ns (pid, &init_pid_ns); -if (p) { -kdb_set_current_task (p); +if (p)return kdb_bt1 (p, ~0UL , false); - }kdb_printf ("No process with pid == %ld found\n" , pid);return 0 ; }else if (strcmp (argv[0 ],"btt" ) ==0 ) {
@@ -178,11 +175,9 @@ kdb_bt(int argc, const char **argv) diag =kdbgetularg ((char *)argv[1 ], &addr);if (diag)return diag; -kdb_set_current_task ((struct task_struct *)addr);return kdb_bt1 ((struct task_struct *)addr, ~0UL , false); }else if (strcmp (argv[0 ],"btc" ) ==0 ) {unsigned long cpu = ~0 ; -struct task_struct *save_current_task = kdb_current_task;if (argc >1 )return KDB_ARGCOUNT ;if (argc ==1 ) {
@@ -204,7 +199,6 @@ kdb_bt(int argc, const char **argv)kdb_bt_cpu (cpu);touch_nmi_watchdog (); } -kdb_set_current_task (save_current_task); }return 0 ; }else {
kernel/debug/kdb/kdb_main.c +1 -1
@@ -1138,7 +1138,7 @@ static void kdb_dumpregs(struct pt_regs *regs) console_loglevel = old_lvl; } -void kdb_set_current_task (struct task_struct *p) +static void kdb_set_current_task (struct task_struct *p) { kdb_current_task = p;
kernel/debug/kdb/kdb_private.h +0 -1
@@ -240,7 +240,6 @@ extern void *debug_kmalloc(size_t size, gfp_t flags);extern void debug_kfree (void *);extern void debug_kusage (void ); -extern void kdb_set_current_task (struct task_struct *);extern struct task_struct *kdb_current_task;extern struct pt_regs *kdb_current_regs;
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
Douglas Anderson
reviewerPATCH 5/5 kdb: Get rid of confusing diag msg from "rd" if current task has no regs
In reply to Douglas AndersonIf you switch to a sleeping task with the "pid" command and then type "rd", kdb tells you this:
No current kdb registers. You may need to select another task
diag: -17: Invalid register name
The first message makes sense, but not the second. Fix it by just returning 0 after commands accessing the current registers finish if we've already printed the "No current kdb registers" error.
While fixing kdb_rd(), change the function to use "if" rather than "ifdef". It cleans the function up a bit and any modern compiler will have no trouble handling still producing good code.
---
kernel/debug/kdb/kdb_main.c | 28 +++++++++++++--------------- 1 file changed, 13 insertions(+), 15 deletions(-)
kernel/debug/kdb/kdb_main.c +13 -15
@@ -543,9 +543,8 @@ int kdbgetaddrarg(int argc, const char **argv, int *nextarg,if (diag)return diag; }else if (symname[0 ] =='%' ) { - diag =kdb_check_regs (); -if (diag) -return diag; +if (kdb_check_regs ()) +return 0 ;/* Implement register values with % at a later time as it is * arch optional. */
@@ -1836,8 +1835,7 @@ static int kdb_go(int argc, const char **argv)*/ static int kdb_rd (int argc,const char **argv) { -int len =kdb_check_regs (); -#if DBG_MAX_REG_NUM >0 +int len =0 ;int i;char *rname;int rsize;
@@ -1846,8 +1844,14 @@ static int kdb_rd(int argc, const char **argv)u16 reg16;u8 reg8; -if (len) -return len; +if (kdb_check_regs ()) +return 0 ; + +/* Fallback to Linux showregs() if we don't have DBG_MAX_REG_NUM */ +if (DBG_MAX_REG_NUM <=0 ) { +kdb_dumpregs (kdb_current_regs); +return 0 ; + }for (i =0 ; i <DBG_MAX_REG_NUM ; i++) { rsize = dbg_reg_def[i].size *2 ;
@@ -1889,12 +1893,7 @@ static int kdb_rd(int argc, const char **argv) } }kdb_printf ("\n" ); -#else -if (len) -return len; -kdb_dumpregs (kdb_current_regs); -#endif return 0 ; }
@@ -1928,9 +1927,8 @@ static int kdb_rm(int argc, const char **argv)if (diag)return diag; - diag =kdb_check_regs (); -if (diag) -return diag; +if (kdb_check_regs ()) +return 0 ; diag =KDB_BADREG ;for (i =0 ; i <DBG_MAX_REG_NUM ; i++) {
Signature · 1 line
2.24.0.432.g9d3f5f5b63-goog
Doug Anderson
reviewerHi
Quoted 17 lines · Douglas Anderson
This started out as just a follow-up to Daniel's post [1]. I wanted to stop implicitly changing the current task in kdb. ...but, of course, everywhere you look in kdb there are things to cleanup, so this gets a few misc cleanups I found along the way. Enjoy.
[1] https://lore.kernel.org/r/[email protected]
Douglas Anderson (5):
MIPS: kdb: Remove old workaround for backtracing on other CPUs
kdb:kdb_current_regsshould be private
kdb:kdb_current_taskshouldn't be exported
kdb: Gid rid of implicit setting of the current task / regs
kdb: Get rid of confusing diag msg from "rd" if current task has no
regs
Quoted 6 lines
arch/mips/kernel/traps.c | 5 ----- include/linux/kdb.h | 2 -- kernel/debug/kdb/kdb_bt.c | 8 +------- kernel/debug/kdb/kdb_main.c | 31 ++++++++++++++----------------- kernel/debug/kdb/kdb_private.h | 2 +- 5 files changed, 16 insertions(+), 32 deletions(-)
I noticed that this series doesn't seem to be in linux-next, but I think it was supposed to target v5.6? Do you know if there is anything outstanding or if it'll be queued up sometime soon?
Thanks!
-Doug
Quoted 27 lines · On Tue, 28 Jan 2020 08:59:01 -0800 Doug Anderson <[email protected]> wrote:
Hi
Quoted 17 lines · Douglas Anderson
This started out as just a follow-up to Daniel's post [1]. I wanted to stop implicitly changing the current task in kdb. ...but, of course, everywhere you look in kdb there are things to cleanup, so this gets a few misc cleanups I found along the way. Enjoy.
[1] https://lore.kernel.org/r/[email protected]
Douglas Anderson (5):
MIPS: kdb: Remove old workaround for backtracing on other CPUs
kdb:kdb_current_regsshould be private
kdb:kdb_current_taskshouldn't be exported
kdb: Gid rid of implicit setting of the current task / regs
kdb: Get rid of confusing diag msg from "rd" if current task has no
regsQuoted 6 lines
arch/mips/kernel/traps.c | 5 ----- include/linux/kdb.h | 2 -- kernel/debug/kdb/kdb_bt.c | 8 +------- kernel/debug/kdb/kdb_main.c | 31 ++++++++++++++----------------- kernel/debug/kdb/kdb_private.h | 2 +- 5 files changed, 16 insertions(+), 32 deletions(-)
Quoted 4 lines
I noticed that this series doesn't seem to be in linux-next, but I think it was supposed to target v5.6? Do you know if there is anything outstanding or if it'll be queued up sometime soon?
I grabbed them.
Are there any updates on the testing status, particularly on MIPS?
Quoted 29 lines · Andrew Morton
Quoted 27 lines · On Tue, 28 Jan 2020 08:59:01 -0800 Doug Anderson <[email protected]> wrote:
Hi
Quoted 23 lines · Douglas Anderson
This started out as just a follow-up to Daniel's post [1]. I wanted to stop implicitly changing the current task in kdb. ...but, of course, everywhere you look in kdb there are things to cleanup, so this gets a few misc cleanups I found along the way. Enjoy.
[1] https://lore.kernel.org/r/[email protected]
Douglas Anderson (5):
MIPS: kdb: Remove old workaround for backtracing on other CPUs
kdb:kdb_current_regsshould be private
kdb:kdb_current_taskshouldn't be exported
kdb: Gid rid of implicit setting of the current task / regs
kdb: Get rid of confusing diag msg from "rd" if current task has no
regsarch/mips/kernel/traps.c | 5 ----- include/linux/kdb.h | 2 -- kernel/debug/kdb/kdb_bt.c | 8 +------- kernel/debug/kdb/kdb_main.c | 31 ++++++++++++++----------------- kernel/debug/kdb/kdb_private.h | 2 +- 5 files changed, 16 insertions(+), 32 deletions(-)
Quoted 8 lines
Quoted 4 lines
I noticed that this series doesn't seem to be in linux-next, but I think it was supposed to target v5.6? Do you know if there is anything outstanding or if it'll be queued up sometime soon?
I grabbed them.
Are there any updates on the testing status, particularly on MIPS?
I put in a fair bit of work earlier this dev cycle to add MIPS to the kgdb test suite. Unfortunately that left me catching up on a few other things... or putting it another way I am late putting together the kgdb tree for v5.6. It will be done by the weekend.
The first patch never got a formal Acked-by from the MIPS folks but they commented positively so we can probably go ahead.
Daniel.