mm: fix kernel crash in khugepaged thread
Subsystems (3)
PATCH v4 1/1 mm: fix kernel crash in khugepaged thread
This crash is caused by NULL pointer deference, in page_to_pfn() marco, when page == NULL :
[ 182.639154 ] Unable to handle kernel NULL pointer dereference at virtual address 00000000 [ 182.639491 ] pgd = ffffffc00077a000 [ 182.639761 ] [00000000] *pgd=00000000b9422003, *pud=00000000b9422003, *pmd=00000000b9423003, *pte=0060000008000707 [ 182.640749 ] Internal error: Oops: 94000006 [#1] SMP [ 182.641197 ] Modules linked in: [ 182.641580 ] CPU: 1 PID: 26 Comm: khugepaged Tainted: G W 4.3.0-rc6-next-20151022ajb-00001-g32f3386-dirty #3 [ 182.642077 ] Hardware name: linux,dummy-virt (DT) [ 182.642227 ] task: ffffffc07957c080 ti: ffffffc079638000 task.ti: ffffffc079638000 [ 182.642598 ] PC is at khugepaged+0x378/0x1af8 [ 182.642826 ] LR is at khugepaged+0x418/0x1af8 [ 182.643047 ] pc : [<ffffffc0001980ac>] lr : [<ffffffc00019814c>] pstate: 60000145 [ 182.643490 ] sp : ffffffc07963bca0 [ 182.643650 ] x29: ffffffc07963bca0 x28: ffffffc00075c000 [ 182.644024 ] x27: ffffffc00f275040 x26: ffffffc0006c7000 [ 182.644334 ] x25: 00e8000048800f51 x24: 0000000006400000 [ 182.644687 ] x23: 0000000000000002 x22: 0000000000000000 [ 182.644972 ] x21: 0000000000000000 x20: 0000000000000000 [ 182.645446 ] x19: 0000000000000000 x18: 0000007ff86d0990 [ 182.645931 ] x17: 00000000007ef9c8 x16: ffffffc000098390 [ 182.646236 ] x15: ffffffffffffffff x14: 00000000ffffffff [ 182.646649 ] x13: 000000000000016a x12: 0000000000000000 [ 182.647046 ] x11: ffffffc07f025020 x10: 0000000000000000 [ 182.647395 ] x9 : 0000000000000048 x8 : ffffffc000721e28 [ 182.647872 ] x7 : 0000000000000000 x6 : ffffffc07f02d000 [ 182.648261 ] x5 : fffffffffffffe00 x4 : ffffffc00f275040 [ 182.648611 ] x3 : 0000000000000000 x2 : ffffffc00f2ad000 [ 182.648908 ] x1 : 0000000000000000 x0 : ffffffc000727000 [ 182.649147 ] [ 182.649252 ] Process khugepaged (pid: 26, stack limit = 0xffffffc079638020) [ 182.649724 ] Stack: (0xffffffc07963bca0 to 0xffffffc07963c000) [ 182.650141 ] bca0: ffffffc07963be30 ffffffc0000b5044 ffffffc07961fb80 ffffffc00072e630 [ 182.650587 ] bcc0: ffffffc0005d5090 0000000000000000 ffffffc000197d34 0000000000000000 [ 182.651009 ] bce0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.651446 ] bd00: ffffffc07963bd90 ffffffc07f1cbf80 000000004f3be003 ffffffc00f2750a4 [ 182.651956 ] bd20: ffffffc00f3bf000 ffffffc000000001 0000000000000001 ffffffc07f085740 [ 182.652520 ] bd40: ffffffc00f2ad188 ffffffc000000000 0000000006200000 ffffffc00f275040 [ 182.652972 ] bd60: ffffffc0006b1a90 ffffffc079638000 ffffffc07963be20 ffffffc00f0144d0 [ 182.653357 ] bd80: ffffffc000000000 0000000006400000 ffffffc00f0144d0 00000a0800000001 [ 182.653793 ] bda0: 0000100000000001 ffffffc000000001 ffffffc07f025000 ffffffc00f2750a8 [ 182.654226 ] bdc0: 00000001000005f8 ffffffc00075a000 0000000006a00000 ffffffc000727000 [ 182.654522 ] bde0: ffffffc0006e8478 ffffffc000000000 0000000100000000 ffffffc078fb9000 [ 182.654869 ] be00: ffffffc07963be30 ffffffc000000000 ffffffc07957c080 ffffffc0000cfc4c [ 182.655225 ] be20: ffffffc07963be20 ffffffc07963be20 0000000000000000 ffffffc000085c50 [ 182.655588 ] be40: ffffffc0000b4f64 ffffffc07961fb80 0000000000000000 0000000000000000 [ 182.656138 ] be60: 0000000000000000 ffffffc0000bee2c ffffffc0000b4f64 0000000000000000 [ 182.656609 ] be80: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.657145 ] bea0: ffffffc07963bea0 ffffffc07963bea0 0000000000000000 ffffffc000000000 [ 182.657475 ] bec0: ffffffc07963bec0 ffffffc07963bec0 0000000000000000 0000000000000000 [ 182.657922 ] bee0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.658558 ] bf00: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.658972 ] bf20: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.659291 ] bf40: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.659722 ] bf60: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.660122 ] bf80: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.660654 ] bfa0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.661064 ] bfc0: 0000000000000000 0000000000000000 0000000000000000 0000000000000005 [ 182.661466 ] bfe0: 0000000000000000 0000000000000000 0000000000000000 0000000000000000 [ 182.661848 ] Call trace: [ 182.662050 ] [<ffffffc0001980ac>] khugepaged+0x378/0x1af8 [ 182.662294 ] [<ffffffc0000b5040>] kthread+0xdc/0xf4 [ 182.662605 ] [<ffffffc000085c4c>] ret_from_fork+0xc/0x40 [ 182.663046 ] Code: 35001700 f0002c60 aa0703e3 f9009fa0 (f94000e0) [ 182.663901 ] ---[ end trace 637503d8e28ae69e ]--- [ 182.664160 ] Kernel panic - not syncing: Fatal exception [ 182.664571 ] CPU2: stopping [ 182.664794 ] CPU: 2 PID: 0 Comm: swapper/2 Tainted: G D W 4.3.0-rc6-next-20151022ajb-00001-g32f3386-dirty #3 [ 182.665248 ] Hardware name: linux,dummy-virt (DT)
add the trace point with TP_CONDITION(page),
avoid trace NULL page.
--- include/trace/events/huge_memory.h | 20 ++++++++++++-------- mm/huge_memory.c | 6 +++--- 2 files changed, 15 insertions(+), 11 deletions(-)
include/trace/events/huge_memory.h +12 -8
@@ -45,12 +45,14 @@ SCAN_STATUS#define EM (a, b) {a, b},#define EMe (a, b) {a, b} -TRACE_EVENT(mm_khugepaged_scan_pmd, +TRACE_EVENT_CONDITION(mm_khugepaged_scan_pmd, - TP_PROTO(struct mm_struct *mm, unsigned long pfn, bool writable, + TP_PROTO(struct mm_struct *mm, struct page *page, bool writable, bool referenced, int none_or_zero, int status, int unmapped), - TP_ARGS(mm, pfn, writable, referenced, none_or_zero, status, unmapped), + TP_ARGS(mm, page, writable, referenced, none_or_zero, status, unmapped), + + TP_CONDITION(page), TP_STRUCT__entry( __field(struct mm_struct *, mm)
@@ -64,7 +66,7 @@ TRACE_EVENT(mm_khugepaged_scan_pmd,TP_fast_assign ( __entry->mm = mm; - __entry->pfn = pfn; + __entry->pfn =page_to_pfn (page); __entry->writable = writable; __entry->referenced = referenced; __entry->none_or_zero = none_or_zero;
@@ -106,12 +108,14 @@ TRACE_EVENT(mm_collapse_huge_page, __print_symbolic(__entry->status, SCAN_STATUS)) ); -TRACE_EVENT(mm_collapse_huge_page_isolate, +TRACE_EVENT_CONDITION(mm_collapse_huge_page_isolate, - TP_PROTO(unsigned long pfn, int none_or_zero, + TP_PROTO(struct page *page, int none_or_zero, bool referenced, bool writable, int status), - TP_ARGS(pfn, none_or_zero, referenced, writable, status), + TP_ARGS(page, none_or_zero, referenced, writable, status), + + TP_CONDITION(page), TP_STRUCT__entry( __field(unsigned long, pfn)
@@ -122,7 +126,7 @@ TRACE_EVENT(mm_collapse_huge_page_isolate, ), TP_fast_assign( - __entry->pfn = pfn; + __entry->pfn = page_to_pfn(page); __entry->none_or_zero = none_or_zero; __entry->referenced = referenced; __entry->writable = writable;
mm/huge_memory.c +3 -3
@@ -1977,7 +1977,7 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,if (likely (writable)) {if (likely (referenced)) { result =SCAN_SUCCEED ; -trace_mm_collapse_huge_page_isolate (page_to_pfn (page), none_or_zero, +trace_mm_collapse_huge_page_isolate (page, none_or_zero, referenced, writable, result);return 1 ; }
@@ -1987,7 +1987,7 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma, out:release_pte_pages (pte, _pte); -trace_mm_collapse_huge_page_isolate (page_to_pfn (page), none_or_zero, +trace_mm_collapse_huge_page_isolate (page, none_or_zero, referenced, writable, result);return 0 ; }
@@ -2530,7 +2530,7 @@ out_unmap:collapse_huge_page (mm, address, hpage, vma, node); } out: -trace_mm_khugepaged_scan_pmd (mm,page_to_pfn (page), writable, referenced, +trace_mm_khugepaged_scan_pmd (mm, page, writable, referenced, none_or_zero, result, unmapped);return ret; }
Signature · 7 lines
1.9.1 -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Steven Rostedt
maintainerQuoted 7 lines · On Thu, 12 Nov 2015 16:21:02 +0800 yalin wang <[email protected]> wrote:
This crash is caused by NULL pointer deference, in
page_to_pfn()marco, when page == NULL :[ 182.639154 ] Unable to handle kernel NULL pointer dereference at virtual address 00000000
add the trace point with
TP_CONDITION(page),
I wonder if we still want to trace even if page is NULL?
Quoted 31 lines
avoid trace NULL page.
Signed-off-by yalin wang <[email protected]>--- include/trace/events/huge_memory.h | 20 ++++++++++++-------- mm/huge_memory.c | 6 +++--- 2 files changed, 15 insertions(+), 11 deletions(-)include/trace/events/huge_memory.h +5
-3@@ -45,12 +45,14 @@ SCAN_STATUS#define EM (a, b) {a, b},#define EMe (a, b) {a, b} -TRACE_EVENT(mm_khugepaged_scan_pmd, +TRACE_EVENT_CONDITION(mm_khugepaged_scan_pmd, - TP_PROTO(struct mm_struct *mm, unsigned long pfn, bool writable, + TP_PROTO(struct mm_struct *mm, struct page *page, bool writable, bool referenced, int none_or_zero, int status, int unmapped), - TP_ARGS(mm, pfn, writable, referenced, none_or_zero, status, unmapped), + TP_ARGS(mm, page, writable, referenced, none_or_zero, status, unmapped), + + TP_CONDITION(page), TP_STRUCT__entry( __field(struct mm_struct *, mm)
include/trace/events/huge_memory.h TRACE_EVENT(mm_khugepaged_scan_pmd, · line 69
Quoted 5 lines
include/trace/events/huge_memory.h +1
-1@@ -64,7 +66,7 @@ TRACE_EVENT(mm_khugepaged_scan_pmd,TP_fast_assign ( __entry->mm = mm; - __entry->pfn = pfn; + __entry->pfn =page_to_pfn (page);
Instead of the condition, we could have:
__entry->pfn = page ? page_to_pfn(page) : -1;
But if there's no reason to do the tracepoint if page is NULL, then this patch is fine. I'm just throwing out this idea.
-- Steve
Quoted 11 lines
include/trace/events/huge_memory.h +1
-2__entry->writable = writable; __entry->referenced = referenced; __entry->none_or_zero = none_or_zero;@@ -106,12 +108,14 @@ TRACE_EVENT(mm_collapse_huge_page, __print_symbolic(__entry->status, SCAN_STATUS)) ); -TRACE_EVENT(mm_collapse_huge_page_isolate, +TRACE_EVENT_CONDITION(mm_collapse_huge_page_isolate, - TP_PROTO(unsigned long pfn, int none_or_zero,
include/trace/events/huge_memory.h TRACE_EVENT(mm_collapse_huge_page, · line 121
Quoted 10 lines
include/trace/events/huge_memory.h +4
-1+ TP_PROTO(struct page *page, int none_or_zero, bool referenced, bool writable, int status), -TP_ARGS (pfn ,none_or_zero ,referenced ,writable ,status ), +TP_ARGS (page, none_or_zero, referenced, writable, status), + +TP_CONDITION (page),TP_STRUCT__entry ( __field(unsigned long, pfn)
\
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 6 lines · Steven Rostedt
Quoted 3 lines · On Thu, 12 Nov 2015 16:21:02 +0800 yalin wang <[email protected]> wrote:
This crash is caused by NULL pointer deference, in
page_to_pfn()marco, when page == NULL :
Quoted 6 lines
Quoted 4 lines
[ 182.639154 ] Unable to handle kernel NULL pointer dereference at virtual address 00000000
add the trace point with
TP_CONDITION(page),I wonder if we still want to trace even if page is NULL?
I'd say we want to. There's even a "SCAN_PAGE_NULL" result defined for that case, and otherwise we would only have to guess why collapsing failed, which is the thing that the tracepoint should help us find out in the first place :)
Quoted 32 lines
Quoted 32 lines
avoid trace NULL page.
Signed-off-by yalin wang <[email protected]>--- include/trace/events/huge_memory.h | 20 ++++++++++++-------- mm/huge_memory.c | 6 +++--- 2 files changed, 15 insertions(+), 11 deletions(-)include/trace/events/huge_memory.h +5
-3@@ -45,12 +45,14 @@ SCAN_STATUS#define EM (a, b) {a, b},#define EMe (a, b) {a, b} -TRACE_EVENT(mm_khugepaged_scan_pmd, +TRACE_EVENT_CONDITION(mm_khugepaged_scan_pmd, - TP_PROTO(struct mm_struct *mm, unsigned long pfn, bool writable, + TP_PROTO(struct mm_struct *mm, struct page *page, bool writable, bool referenced, int none_or_zero, int status, int unmapped), - TP_ARGS(mm, pfn, writable, referenced, none_or_zero, status, unmapped), + TP_ARGS(mm, page, writable, referenced, none_or_zero, status, unmapped), + + TP_CONDITION(page), TP_STRUCT__entry( __field(struct mm_struct *, mm)@@ -64,7 +66,7 @@ TRACE_EVENT(mm_khugepaged_scan_pmd,
Quoted 8 lines
include/trace/events/huge_memory.h TRACE_EVENT(mm_khugepaged_scan_pmd, · line 69
Quoted 4 lines
include/trace/events/huge_memory.h +1
-1TP_fast_assign ( __entry->mm = mm; - __entry->pfn = pfn; + __entry->pfn =page_to_pfn (page);Instead of the condition, we could have:
__entry->pfn = page ? page_to_pfn(page) : -1;
I agree. Please do it like this.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 58 lines · Vlastimil Babka
Quoted 6 lines · Steven Rostedt
Quoted 3 lines · On Thu, 12 Nov 2015 16:21:02 +0800 yalin wang <[email protected]> wrote:
This crash is caused by NULL pointer deference, in
page_to_pfn()marco, when page == NULL :Quoted 6 lines
Quoted 4 lines
[ 182.639154 ] Unable to handle kernel NULL pointer dereference at virtual address 00000000
add the trace point with
TP_CONDITION(page),I wonder if we still want to trace even if page is NULL?
I'd say we want to. There's even a "
SCAN_PAGE_NULL" result defined for that case, and otherwise we would only have to guess why collapsing failed, which is the thing that the tracepoint should help us find out in the first place :)Quoted 39 lines
Quoted 31 lines
avoid trace NULL page.
Signed-off-by yalin wang <[email protected]>--- include/trace/events/huge_memory.h | 20 ++++++++++++-------- mm/huge_memory.c | 6 +++--- 2 files changed, 15 insertions(+), 11 deletions(-)include/trace/events/huge_memory.h +5
-3@@ -45,12 +45,14 @@ SCAN_STATUS#define EM (a, b) {a, b},#define EMe (a, b) {a, b} -TRACE_EVENT(mm_khugepaged_scan_pmd, +TRACE_EVENT_CONDITION(mm_khugepaged_scan_pmd, - TP_PROTO(struct mm_struct *mm, unsigned long pfn, bool writable, + TP_PROTO(struct mm_struct *mm, struct page *page, bool writable, bool referenced, int none_or_zero, int status, int unmapped), - TP_ARGS(mm, pfn, writable, referenced, none_or_zero, status, unmapped), + TP_ARGS(mm, page, writable, referenced, none_or_zero, status, unmapped), + + TP_CONDITION(page), TP_STRUCT__entry( __field(struct mm_struct *, mm)include/trace/events/huge_memory.h TRACE_EVENT(mm_khugepaged_scan_pmd, · line 69
Quoted 5 lines
include/trace/events/huge_memory.h +1
-1@@ -64,7 +66,7 @@ TRACE_EVENT(mm_khugepaged_scan_pmd,TP_fast_assign ( __entry->mm = mm; - __entry->pfn = pfn; + __entry->pfn =page_to_pfn (page);Instead of the condition, we could have:
Quoted 3 lines
Quoted 1 line
__entry->pfn = page ? page_to_pfn(page) : -1;I agree. Please do it like this.
ok , i will send V5 patch .
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Steven Rostedt
maintainerQuoted 10 lines · On Fri, 13 Nov 2015 19:54:11 +0800 yalin wang <[email protected]> wrote:
Quoted 10 lines
Quoted 8 lines
include/trace/events/huge_memory.h TRACE_EVENT(mm_khugepaged_scan_pmd, · line 69
Quoted 4 lines
include/trace/events/huge_memory.h +1
-1TP_fast_assign ( __entry->mm = mm; - __entry->pfn = pfn; + __entry->pfn =page_to_pfn (page);Instead of the condition, we could have:
__entry->pfn = page ? page_to_pfn(page) : -1;I agree. Please do it like this.
hmm, pfn is defined as an unsigned long, would -1 be the best. Or should it be (-1UL).
Then we could also have:
TP_printk("mm=%p, scan_pfn=0x%lx%s, writable=%d, referenced=%d, none_or_zero=%d, status=%s, unmapped=%d",
__entry->mm,
__entry->pfn == (-1UL) ? 0 : __entry->pfn,
__entry->pfn == (-1UL) ? "(null)" : "",Note the added %s after %lx I have in the print format.
-- Steve
Quoted 1 line
ok , i will send V5 patch .
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 22 lines · Steven Rostedt
Quoted 10 lines · On Fri, 13 Nov 2015 19:54:11 +0800 yalin wang <[email protected]> wrote:
Quoted 10 lines
Quoted 8 lines
Review
Quoted 4 lines
TP_fast_assign( __entry->mm = mm;(quoted diff) +1
-1- __entry->pfn = pfn; + __entry->pfn = page_to_pfn(page);Instead of the condition, we could have:
__entry->pfn = page ? page_to_pfn(page) : -1;I agree. Please do it like this.
hmm, pfn is defined as an unsigned long, would -1 be the best. Or should it be (-1UL).
Then we could also have:
Quoted 8 lines
TP_printk("mm=%p, scan_pfn=0x%lx%s, writable=%d, referenced=%d, none_or_zero=%d, status=%s, unmapped=%d", __entry->mm, __entry->pfn == (-1UL) ? 0 : __entry->pfn, __entry->pfn == (-1UL) ? "(null)" : "",Note the added %s after %lx I have in the print format.
-- Steve
it is not easy to print for perf tools in userspace ,
if you use this format ,
for user space perf tool, it print the entry by look up the member in entry struct by offset , you print a dynamic string which user space perf tool don’t know how to print this string .
Thanks
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 27 lines · yalin wang
Quoted 27 lines · Steven Rostedt
Quoted 10 lines · On Fri, 13 Nov 2015 19:54:11 +0800 yalin wang <[email protected]> wrote:
Quoted 10 lines
Quoted 8 lines
Review
Quoted 4 lines
TP_fast_assign( __entry->mm = mm;(quoted diff) +1
-1- __entry->pfn = pfn; + __entry->pfn = page_to_pfn(page);Instead of the condition, we could have:
__entry->pfn = page ? page_to_pfn(page) : -1;I agree. Please do it like this.
hmm, pfn is defined as an unsigned long, would -1 be the best. Or should it be (-1UL).
Then we could also have:
TP_printk("mm=%p, scan_pfn=0x%lx%s, writable=%d, referenced=%d, none_or_zero=%d, status=%s, unmapped=%d", __entry->mm, __entry->pfn == (-1UL) ? 0 : __entry->pfn, __entry->pfn == (-1UL) ? "(null)" : "",
Quoted 7 lines
Quoted 3 lines
Note the added %s after %lx I have in the print format.
-- Steve
it is not easy to print for perf tools in userspace ,
if you use this format ,
for user space perf tool, it print the entry by look up the member in entry struct by offset , you print a dynamic string which user space perf tool dona??t know how to print this string .
Does it work through trace-cmd?
Quoted 2 lines
Thanks
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Steven Rostedt
maintainerQuoted 8 lines · On Mon, 16 Nov 2015 11:16:22 +0100 Vlastimil Babka <[email protected]> wrote:
Quoted 6 lines
Quoted 2 lines
-- Steve
it is not easy to print for perf tools in userspace ,
if you use this format ,
for user space perf tool, it print the entry by look up the member in entry struct by offset , you print a dynamic string which user space perf tool don’t know how to print this string .Does it work through trace-cmd?
The two use the same code. If it works in one, it will work in the other.
-- Steve
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 12 lines · Steven Rostedt
Quoted 7 lines · On Mon, 16 Nov 2015 11:16:22 +0100 Vlastimil Babka <[email protected]> wrote:
Quoted 6 lines
Quoted 2 lines
-- Steve
it is not easy to print for perf tools in userspace ,
if you use this format ,
for user space perf tool, it print the entry by look up the member in entry struct by offset , you print a dynamic string which user space perf tool don’t know how to print this string .
Quoted 7 lines
Quoted 1 line
Does it work through trace-cmd?
The two use the same code. If it works in one, it will work in the other.
-- Steve
i have not tried ,
just a question,
if you print a %s , but don’t call trace_define_field() do define this string in __entry , how does user space perf tool to get this string info and print it ? i am curious ..
i can try this when i have time. and report to you .
Thanks
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Steven Rostedt
maintainerQuoted 6 lines · On Tue, 17 Nov 2015 10:21:47 +0800 yalin wang <[email protected]> wrote:
i have not tried ,
just a question,
if you print a %s , but don’t calltrace_define_field()do define this string in__entry, how does user space perf tool to get this string info and print it ? i am curious ..
i can try this when i have time. and report to you .
Because the print_fmt has nothing to do with the fields. You can have as your print_fmt as:
TP_printk("Message = %s", "hello dolly!")And both userspace and the kernel with process that correctly (if I got string processing working in userspace, which I believe I do). The string is processed, it's not dependent on TP_STRUCT__entry() unless it references a field there. Which can also be used too:
TP_printk("Message = %s", __entry->musical ? "Hello dolly!" :
"Death Trap!")userspace will see in the entry:
print_fmt: "Message = %s", REC->musical ? "Hello dolly!" : "Death Trap!"
as long as the field "musical" exists, all is well.
-- Steve
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 27 lines · Steven Rostedt
Quoted 6 lines · On Tue, 17 Nov 2015 10:21:47 +0800 yalin wang <[email protected]> wrote:
i have not tried ,
just a question,
if you print a %s , but don’t calltrace_define_field()do define this string in__entry, how does user space perf tool to get this string info and print it ? i am curious ..
i can try this when i have time. and report to you .Because the
print_fmthas nothing to do with the fields. You can have as yourprint_fmtas:TP_printk("Message = %s", "hello dolly!")And both userspace and the kernel with process that correctly (if I got string processing working in userspace, which I believe I do). The string is processed, it's not dependent on
TP_STRUCT__entry()unless it references a field there. Which can also be used too:TP_printk("Message = %s", __entry->musical ? "Hello dolly!" : "Death Trap!")
Quoted 7 lines
userspace will see in the entry:
print_fmt: "Message = %s",REC->musical? "Hello dolly!" : "Death Trap!"as long as the field "musical" exists, all is well.
-- Steve
Aha, i see.
Thanks very much for your explanation.
Better print fat is :
TP_printk("mm=%p, scan_pfn=%s, writable=%d, referenced=%d, none_or_zero=%d, status=%s, unmapped=%d",
__entry->mm, __entry->pfn == (-1UL) ? "(null)" : itoa(buff, __entry->pin, 10), …..)
is this possible ?
Thanks
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Quoted 28 lines · yalin wang
Quoted 21 lines · Steven Rostedt
On Tue, 17 Nov 2015 10:21:47 +0800
yalin wang <[email protected]> wrote:Because the
print_fmthas nothing to do with the fields. You can have as yourprint_fmtas:TP_printk("Message = %s", "hello dolly!")And both userspace and the kernel with process that correctly (if I got string processing working in userspace, which I believe I do). The string is processed, it's not dependent on
TP_STRUCT__entry()unless it references a field there. Which can also be used too:TP_printk("Message = %s", __entry->musical ? "Hello dolly!" : "Death Trap!")Quoted 7 lines
userspace will see in the entry:
print_fmt: "Message = %s",REC->musical? "Hello dolly!" : "Death Trap!"as long as the field "musical" exists, all is well.
-- Steve
Quoted 8 lines
Aha, i see.
Thanks very much for your explanation.
Better print fat is :
TP_printk("mm=%p, scan_pfn=%s, writable=%d, referenced=%d, none_or_zero=%d, status=%s, unmapped=%d",__entry->mm, __entry->pfn == (-1UL) ? "(null)" : itoa(buff, __entry->pin, 10), a?|..)is this possible ?
I doubt so.
Why don't we just do (with %lx):
__entry->pfn != -1UL ? __entry->pfn : 0,
Status already tells us that it's not a real pfn 0 (which I doubt would be userspace-mapped and thus reachable by khugepaged anyway?). Also it's what some other tracepoints do, see e.g. mm_page class in include/trace/events/kmem.h.
Quoted 9 lines
Thanks
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>
Steven Rostedt
maintainerQuoted 26 lines · On Mon, 16 Nov 2015 09:35:53 +0800 yalin wang <[email protected]> wrote:
Quoted 26 lines · Steven Rostedt
Quoted 10 lines · On Fri, 13 Nov 2015 19:54:11 +0800 yalin wang <[email protected]> wrote:
Quoted 10 lines
Quoted 8 lines
Review
Quoted 4 lines
TP_fast_assign( __entry->mm = mm;(quoted diff) +1
-1- __entry->pfn = pfn; + __entry->pfn = page_to_pfn(page);Instead of the condition, we could have:
__entry->pfn = page ? page_to_pfn(page) : -1;I agree. Please do it like this.
hmm, pfn is defined as an unsigned long, would -1 be the best. Or should it be (-1UL).
Then we could also have:
TP_printk("mm=%p, scan_pfn=0x%lx%s, writable=%d, referenced=%d, none_or_zero=%d, status=%s, unmapped=%d", __entry->mm, __entry->pfn == (-1UL) ? 0 : __entry->pfn, __entry->pfn == (-1UL) ? "(null)" : "",
Quoted 7 lines
Quoted 3 lines
Note the added %s after %lx I have in the print format.
-- Steve
it is not easy to print for perf tools in userspace ,
if you use this format ,
for user space perf tool, it print the entry by look up the member in entry struct by offset , you print a dynamic string which user space perf tool don’t know how to print this string .
Have you tried it? It should work. If not, I'll fix it. The string "null" is exported in the trace output file, and perf should have enough information to know how to handle that. If it fails to parse, I can easily fix it.
Remember, I'm the author of the parsing of events in userspace.
-- Steve
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to [email protected]. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"[email protected]"> [email protected] </a>