Skip to contentSkip to conversation

xentrace changes

PATCH In review Olaf Hering 3 participants 18 messages Jun 1, 2023 – Aug 1, 2023

lore
Share & tools
lore
#0 lore raw

cover xentrace changes

Olaf Hering (7):
xentrace: remove unimplemented option from man page
xentrace: use correct output format for pit and rtc
xenalyze: sync with vmx.h, use EXIT_REASON_PENDING_VIRT_INTR
xenalyze: sync with vmx.h, use EXIT_REASON_VMXOFF
xenalyze: sync with vmx.h, use EXIT_REASON_VMXON
xenalyze: sync with vmx.h, use EXIT_REASON_MCE_DURING_VMENTRY
xenalyze: handle more potential exit reason values from vmx.h

 docs/man/xentrace.8.pod   |  4 ----
 tools/xentrace/formats    |  4 ++--
 tools/xentrace/xenalyze.c | 46 ++++++++++++++++++++++++++++-----------
 3 files changed, 35 insertions(+), 19 deletions(-)
#1 lore raw

PATCH 1/7 xentrace: remove unimplemented option from man page

The documented option --usage worked because every unknown option showed the help.

---
 docs/man/xentrace.8.pod | 4 ----
 1 file changed, 4 deletions(-)
docs/man/xentrace.8.pod +0 -4
index 69aef05f65..4c174a84c0 100644
@@ -69,10 +69,6 @@ set event capture mask. If not specified the TRC_ALL will be used.
 
 =item B<-?>, B<--help>
 
-Give this help list
-
-=item B<--usage>
-
 Give a short usage message
 
 =item B<-V>, B<--version>
#3 lore raw

PATCH 2/7 xentrace: use correct output format for pit and rtc

In reply to Olaf Hering

The input values were always 32bit.

Fixes 55ee5dea32 ("xentrace: add TRC_HVM_EMUL")

---
 tools/xentrace/formats | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/xentrace/formats +2 -2
index 0fcc327a40..afb5ee0112 100644
@@ -211,8 +211,8 @@
 0x00802008  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  do_irq [ irq = %(1)d, began = %(2)dus, ended = %(3)dus ]
 
 0x00084001  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  hpet create [ tn = %(1)d, irq = %(2)d, delta = 0x%(4)08x%(3)08x, period = 0x%(6)08x%(5)08x ]
-0x00084002  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  pit create [ delta = 0x%(1)016x, period = 0x%(2)016x ]
-0x00084003  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  rtc create [ delta = 0x%(1)016x , period = 0x%(2)016x ]
+0x00084002  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  pit create [ delta = 0x%(1)08x, period = 0x%(2)08x ]
+0x00084003  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  rtc create [ delta = 0x%(1)08x, period = 0x%(2)08x ]
 0x00084004  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  vlapic create [ delta = 0x%(2)08x%(1)08x , period = 0x%(4)08x%(3)08x, irq = %(5)d ]
 0x00084005  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  hpet destroy [ tn = %(1)d ]
 0x00084006  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  pit destroy  [ ]
#4 lore raw

Re: [PATCH v1 2/7] xentrace: use correct output format for pit and rtc

Sorry for the delay here.

Just a couple of notes on the whole series. First, you didn't CC me on 0/7 or 1/7. Secondly, patch 7 seems to be corrupted (see e.g., Patchew's attempt to apply the series [1]).

[1] https://patchew.org/Xen/[email protected]/

-George

Quoted 33 lines · Olaf Hering

The input values were always 32bit.

Fixes 55ee5dea32 ("xentrace: add TRC_HVM_EMUL")

---
 tools/xentrace/formats | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/xentrace/formats +0 -0
index 0fcc327a40..afb5ee0112 100644
@@ -211,8 +211,8 @@
 0x00802008  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  do_irq [ irq = %(1)d,

began = %(2)dus, ended = %(3)dus ]

0x00084001 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) hpet create [ tn = %(1)d,
irq = %(2)d, delta = 0x%(4)08x%(3)08x, period = 0x%(6)08x%(5)08x ] -0x00084002 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) pit create [ delta = 0x%(1)016x, period = 0x%(2)016x ]
-0x00084003 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) rtc create [ delta = 0x%(1)016x , period = 0x%(2)016x ]
+0x00084002 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) pit create [ delta = 0x%(1)08x, period = 0x%(2)08x ]
+0x00084003 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) rtc create [ delta = 0x%(1)08x, period = 0x%(2)08x ]
0x00084004 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) vlapic create [ delta =
0x%(2)08x%(1)08x , period = 0x%(4)08x%(3)08x, irq = %(5)d ]
0x00084005 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) hpet destroy [ tn = %(1)d
]
0x00084006 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) pit destroy [ ]

#5 lore raw

Re: [PATCH v1 2/7] xentrace: use correct output format for pit and rtc

Fri, 28 Jul 2023 17:47:11 +0100 George Dunlap <[email protected]>:

Quoted 3 lines

Just a couple of notes on the whole series. First, you didn't CC me on 0/7 or 1/7. Secondly, patch 7 seems to be corrupted (see e.g., Patchew's attempt to apply the series [1]).

The cover letter can not be easily fixed, this is a bug in git send-email. It should be able to collect all recipients and put them in the cover letter. I think I already pointed that out a while ago on git-devel...

I think #1 can be fixed by adjusting the MAINTAINERS file. Not sure why #7 fails to apply.

I will rebase and resend next week.

Olaf

#6 lore raw

Re: [PATCH v1 2/7] xentrace: use correct output format for pit and rtc

Quoted 7 lines · Olaf Hering

Fri, 28 Jul 2023 17:47:11 +0100 George Dunlap <[email protected]>:

Quoted 1 line

Just a couple of notes on the whole series. First, you didn't CC me on

0/7

Quoted 2 lines

or 1/7. Secondly, patch 7 seems to be corrupted (see e.g., Patchew's attempt to apply the series [1]).

Quoted 8 lines

The cover letter can not be easily fixed, this is a bug in git send-email. It should be able to collect all recipients and put them in the cover letter.
I think I already pointed that out a while ago on git-devel...

I think #1 can be fixed by adjusting the MAINTAINERS file. Not sure why #7 fails to apply.

Running git am --show-current-patch reveals a tail that looks like this:

```
+    [EXIT_REASON_NOTIFY]=3D"NOTIFY",
 };
=20
```

This may again be a known issue with git am and base64-encoded files: I think mailsplit knows how to deal with these sorts of line breaks; and git am also knows how to decode base64; but it does them in the wrong order, so base64-encoded mails don't get handled properly. I reported it to them several years ago [1], and nothing happened.

-George

[1]
https://lore.kernel.org/git/[email protected]/

Reviewed-by on v1 2/7
#7 lore raw

Re: [PATCH v1 2/7] xentrace: use correct output format for pit and rtc

In reply to Olaf Hering
Quoted 17 lines · Olaf Hering

The input values were always 32bit.

Fixes 55ee5dea32 ("xentrace: add TRC_HVM_EMUL")

---
 tools/xentrace/formats | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/xentrace/formats +0 -0
index 0fcc327a40..afb5ee0112 100644
@@ -211,8 +211,8 @@
 0x00802008  CPU%(cpu)d  %(tsc)d (+%(reltsc)8d)  do_irq [ irq = %(1)d,

began = %(2)dus, ended = %(3)dus ]

Quoted 11 lines

0x00084001 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) hpet create [ tn = %(1)d,
irq = %(2)d, delta = 0x%(4)08x%(3)08x, period = 0x%(6)08x%(5)08x ] -0x00084002 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) pit create [ delta = 0x%(1)016x, period = 0x%(2)016x ]
-0x00084003 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) rtc create [ delta = 0x%(1)016x , period = 0x%(2)016x ]
+0x00084002 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) pit create [ delta = 0x%(1)08x, period = 0x%(2)08x ]
+0x00084003 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) rtc create [ delta = 0x%(1)08x, period = 0x%(2)08x ]

Looks good. While we're here, any reason not to s/create/start/; to match up with the trace definition?

Either way:

#8 lore raw

PATCH 3/7 xenalyze: sync with vmx.h, use EXIT_REASON_PENDING_VIRT_INTR

In reply to Olaf Hering
---
 tools/xentrace/xenalyze.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
tools/xentrace/xenalyze.c +3 -3
index a50538e9a8..1cae055ef4 100644
@@ -447,7 +447,7 @@ struct {
 #define EXIT_REASON_SIPI                4
 #define EXIT_REASON_IO_SMI              5
 #define EXIT_REASON_OTHER_SMI           6
-#define EXIT_REASON_PENDING_INTERRUPT   7
+#define EXIT_REASON_PENDING_VIRT_INTR   7
 #define EXIT_REASON_PENDING_VIRT_NMI    8
 #define EXIT_REASON_TASK_SWITCH         9
 #define EXIT_REASON_CPUID               10
@@ -503,7 +503,7 @@ const char * hvm_vmx_exit_reason_name[HVM_VMX_EXIT_REASON_MAX] = {
     [EXIT_REASON_SIPI]="SIPI",
     [EXIT_REASON_IO_SMI]="IO_SMI",
     [EXIT_REASON_OTHER_SMI]="OTHER_SMI",
-    [EXIT_REASON_PENDING_INTERRUPT]="PENDING_INTERRUPT",
+    [EXIT_REASON_PENDING_VIRT_INTR]="PENDING_VIRT_INTR",
     [EXIT_REASON_PENDING_VIRT_NMI]="PENDING_VIRT_NMI",
     [EXIT_REASON_TASK_SWITCH]="TASK_SWITCH",
     [EXIT_REASON_CPUID]="CPUID",
@@ -4632,7 +4632,7 @@ void hvm_generic_postprocess(struct hvm_data *h)
             switch(h->exit_reason)
             {
                 /* These just need us to go through the return path */
-            case EXIT_REASON_PENDING_INTERRUPT:
+            case EXIT_REASON_PENDING_VIRT_INTR:
             case EXIT_REASON_TPR_BELOW_THRESHOLD:
                 /* Not much to log now; may need later */
             case EXIT_REASON_WBINVD:
#10 lore raw

PATCH 4/7 xenalyze: sync with vmx.h, use EXIT_REASON_VMXOFF

In reply to Olaf Hering
---
 tools/xentrace/xenalyze.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/xentrace/xenalyze.c +2 -2
index 1cae055ef4..d2e6c77590 100644
@@ -466,7 +466,7 @@ struct {
 #define EXIT_REASON_VMREAD              23
 #define EXIT_REASON_VMRESUME            24
 #define EXIT_REASON_VMWRITE             25
-#define EXIT_REASON_VMOFF               26
+#define EXIT_REASON_VMXOFF              26
 #define EXIT_REASON_VMON                27
 #define EXIT_REASON_CR_ACCESS           28
 #define EXIT_REASON_DR_ACCESS           29
@@ -522,7 +522,7 @@ const char * hvm_vmx_exit_reason_name[HVM_VMX_EXIT_REASON_MAX] = {
     [EXIT_REASON_VMREAD]="VMREAD",
     [EXIT_REASON_VMRESUME]="VMRESUME",
     [EXIT_REASON_VMWRITE]="VMWRITE",
-    [EXIT_REASON_VMOFF]="VMOFF",
+    [EXIT_REASON_VMXOFF]="VMXOFF",
     [EXIT_REASON_VMON]="VMON",
     [EXIT_REASON_CR_ACCESS]="CR_ACCESS",
     [EXIT_REASON_DR_ACCESS]="DR_ACCESS",
#11 lore raw

PATCH 5/7 xenalyze: sync with vmx.h, use EXIT_REASON_VMXON

In reply to Olaf Hering
---
 tools/xentrace/xenalyze.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/xentrace/xenalyze.c +2 -2
index d2e6c77590..88c3d5f873 100644
@@ -467,7 +467,7 @@ struct {
 #define EXIT_REASON_VMRESUME            24
 #define EXIT_REASON_VMWRITE             25
 #define EXIT_REASON_VMXOFF              26
-#define EXIT_REASON_VMON                27
+#define EXIT_REASON_VMXON               27
 #define EXIT_REASON_CR_ACCESS           28
 #define EXIT_REASON_DR_ACCESS           29
 #define EXIT_REASON_IO_INSTRUCTION      30
@@ -523,7 +523,7 @@ const char * hvm_vmx_exit_reason_name[HVM_VMX_EXIT_REASON_MAX] = {
     [EXIT_REASON_VMRESUME]="VMRESUME",
     [EXIT_REASON_VMWRITE]="VMWRITE",
     [EXIT_REASON_VMXOFF]="VMXOFF",
-    [EXIT_REASON_VMON]="VMON",
+    [EXIT_REASON_VMXON]="VMXON",
     [EXIT_REASON_CR_ACCESS]="CR_ACCESS",
     [EXIT_REASON_DR_ACCESS]="DR_ACCESS",
     [EXIT_REASON_IO_INSTRUCTION]="IO_INSTRUCTION",
#12 lore raw

PATCH 6/7 xenalyze: sync with vmx.h, use EXIT_REASON_MCE_DURING_VMENTRY

In reply to Olaf Hering
---
 tools/xentrace/xenalyze.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
tools/xentrace/xenalyze.c +2 -2
index 88c3d5f873..9635ff453a 100644
@@ -479,7 +479,7 @@ struct {
 #define EXIT_REASON_MONITOR_TRAP_FLAG   37
 #define EXIT_REASON_MONITOR_INSTRUCTION 39
 #define EXIT_REASON_PAUSE_INSTRUCTION   40
-#define EXIT_REASON_MACHINE_CHECK       41
+#define EXIT_REASON_MCE_DURING_VMENTRY  41
 #define EXIT_REASON_TPR_BELOW_THRESHOLD 43
 #define EXIT_REASON_APIC_ACCESS         44
 #define EXIT_REASON_ACCESS_GDTR_OR_IDTR 46
@@ -535,7 +535,7 @@ const char * hvm_vmx_exit_reason_name[HVM_VMX_EXIT_REASON_MAX] = {
     [EXIT_REASON_MONITOR_TRAP_FLAG]="MONITOR_TRAP_FLAG",
     [EXIT_REASON_MONITOR_INSTRUCTION]="MONITOR_INSTRUCTION",
     [EXIT_REASON_PAUSE_INSTRUCTION]="PAUSE_INSTRUCTION",
-    [EXIT_REASON_MACHINE_CHECK]="MACHINE_CHECK",
+    [EXIT_REASON_MCE_DURING_VMENTRY]="MCE_DURING_VMENTRY",
     [EXIT_REASON_TPR_BELOW_THRESHOLD]="TPR_BELOW_THRESHOLD",
     [EXIT_REASON_APIC_ACCESS]="APIC_ACCESS",
     [EXIT_REASON_EPT_VIOLATION]="EPT_VIOLATION",
#15 lore raw

PATCH 7/7 xenalyze: handle more potential exit reason values from vmx.h

In reply to Olaf Hering

Copy and use more constants from vmx.h, to turn numbers into strings. Adjust the REASON_MAX value accordingly.
Remove the size constraint from string array, the compiler will grow it as needed.

---
 tools/xentrace/xenalyze.c | 28 ++++++++++++++++++++++++----
 1 file changed, 24 insertions(+), 4 deletions(-)
tools/xentrace/xenalyze.c +24 -4
index 9635ff453a..9af17d45bf 100644
@@ -482,6 +482,7 @@ struct {
 #define EXIT_REASON_MCE_DURING_VMENTRY  41
 #define EXIT_REASON_TPR_BELOW_THRESHOLD 43
 #define EXIT_REASON_APIC_ACCESS         44
+#define EXIT_REASON_EOI_INDUCED         45
 #define EXIT_REASON_ACCESS_GDTR_OR_IDTR 46
 #define EXIT_REASON_ACCESS_LDTR_OR_TR   47
 #define EXIT_REASON_EPT_VIOLATION       48
@@ -492,10 +493,18 @@ struct {
 #define EXIT_REASON_INVVPID             53
 #define EXIT_REASON_WBINVD              54
 #define EXIT_REASON_XSETBV              55
-
-#define HVM_VMX_EXIT_REASON_MAX (EXIT_REASON_XSETBV+1)
-
-const char * hvm_vmx_exit_reason_name[HVM_VMX_EXIT_REASON_MAX] = {
+#define EXIT_REASON_APIC_WRITE          56
+#define EXIT_REASON_INVPCID             58
+#define EXIT_REASON_VMFUNC              59
+#define EXIT_REASON_PML_FULL            62
+#define EXIT_REASON_XSAVES              63
+#define EXIT_REASON_XRSTORS             64
+#define EXIT_REASON_BUS_LOCK            74
+#define EXIT_REASON_NOTIFY              75
+
+#define HVM_VMX_EXIT_REASON_MAX (EXIT_REASON_NOTIFY+1)
+
+const char * hvm_vmx_exit_reason_name[] = {
     [EXIT_REASON_EXCEPTION_NMI]="EXCEPTION_NMI",
     [EXIT_REASON_EXTERNAL_INTERRUPT]="EXTERNAL_INTERRUPT",
     [EXIT_REASON_TRIPLE_FAULT]="TRIPLE_FAULT",
@@ -538,6 +547,9 @@ const char * hvm_vmx_exit_reason_name[HVM_VMX_EXIT_REASON_MAX] = {
     [EXIT_REASON_MCE_DURING_VMENTRY]="MCE_DURING_VMENTRY",
     [EXIT_REASON_TPR_BELOW_THRESHOLD]="TPR_BELOW_THRESHOLD",
     [EXIT_REASON_APIC_ACCESS]="APIC_ACCESS",
+    [EXIT_REASON_EOI_INDUCED]="EOI_INDUCED",
+    [EXIT_REASON_ACCESS_GDTR_OR_IDTR]="ACCESS_GDTR_OR_IDTR",
+    [EXIT_REASON_ACCESS_LDTR_OR_TR]="ACCESS_LDTR_OR_TR",
     [EXIT_REASON_EPT_VIOLATION]="EPT_VIOLATION",
     [EXIT_REASON_EPT_MISCONFIG]="EPT_MISCONFIG",
     [EXIT_REASON_INVEPT]="INVEPT",
@@ -546,6 +558,14 @@ const char * hvm_vmx_exit_reason_name[HVM_VMX_EXIT_REASON_MAX] = {
     [EXIT_REASON_INVVPID]="INVVPID",
     [EXIT_REASON_WBINVD]="WBINVD",
     [EXIT_REASON_XSETBV]="XSETBV",
+    [EXIT_REASON_APIC_WRITE]="APIC_WRITE",
+    [EXIT_REASON_INVPCID]="INVPCID",
+    [EXIT_REASON_VMFUNC]="VMFUNC",
+    [EXIT_REASON_PML_FULL]="PML_FULL",
+    [EXIT_REASON_XSAVES]="XSAVES",
+    [EXIT_REASON_XRSTORS]="XRSTORS",
+    [EXIT_REASON_BUS_LOCK]="BUS_LOCK",
+    [EXIT_REASON_NOTIFY]="NOTIFY",
 };
 
 /* SVM data */
#16 lore raw

Re: [PATCH v1 7/7] xenalyze: handle more potential exit reason values from vmx.h

Quoted 7 lines · Olaf Hering

Copy and use more constants from vmx.h, to turn numbers into strings. Adjust the REASON_MAX value accordingly.
Remove the size constraint from string array, the compiler will grow it as needed.

Thanks for doing this.

Everything looks good (including adding the missing strings), except for the removal of the fixed array size. Call me paranoid, but if we define it as REASON_MAX length, then there will never be any way that a value less than REASON_MAX causes a segfault (and if we assign a value higher than REASON_MAX, the compiler will complain); whereas if we make it variable, we leave open the possibility that someone won't update REASON_MAX properly, resulting in either segfaults (if REASON_MAX is too high) or missing strings (if REASON_MAX is too low).

-George

#17 lore raw

Re: [PATCH v1 7/7] xenalyze: handle more potential exit reason values from vmx.h

Fri, 28 Jul 2023 21:35:54 +0100 George Dunlap <[email protected]>:

Quoted 8 lines

Everything looks good (including adding the missing strings), except for the removal of the fixed array size. Call me paranoid, but if we define it as REASON_MAX length, then there will never be any way that a value less than REASON_MAX causes a segfault (and if we assign a value higher than REASON_MAX, the compiler will complain); whereas if we make it variable, we leave open the possibility that someone won't update REASON_MAX properly, resulting in either segfaults (if REASON_MAX is too high) or missing strings (if REASON_MAX is too low).

I think the code needs to be rearranged to work with "array_size" instead.

Olaf