Skip to contentSkip to conversation

xen: debugtrace cleanup and per-cpu buffer support

PATCH Superseded Juergen Gross 2 participants 21 messages Aug 28, 2019 – Sep 3, 2019

v3 of 7 Superseded Latest is v7 →

lore
Share & tools
lore
#0 lore raw

v3 cover xen: debugtrace cleanup and per-cpu buffer support

Another debugtrace enhancement I needed for core scheduling debugging, plus some cleanup.

Changes in V3:
- rebase to current staging

Changes in V2:
- added new patch 1 (preparing the move of debugtrace coding)
- patch 4 (v1 patch 3): avoid leaking buffer

Juergen Gross (4):
xen: use common output function in debugtrace
xen: move debugtrace coding to common/debugtrace.c
xen: refactor debugtrace data
xen: add per-cpu buffer option to debugtrace

 docs/misc/xen-command-line.pandoc |   7 +-
 xen/common/Makefile               |   1 +
 xen/common/debugtrace.c           | 265 ++++++++++++++++++++++++++++++++++++++
 xen/drivers/char/console.c        | 186 +-------------------------
 xen/include/xen/console.h         |   3 +
 5 files changed, 278 insertions(+), 184 deletions(-)
 create mode 100644 xen/common/debugtrace.c
Signature · 7 lines
2.16.4


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#1 lore raw

PATCH v3 1/4 xen: use common output function in debugtrace

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

In order to prepare moving debugtrace functionality to an own source file rename sercon_puts() to console_serial_puts() and make it globally visible.

---
 xen/drivers/char/console.c | 18 +++++++++---------
 xen/include/xen/console.h  |  3 +++
 2 files changed, 12 insertions(+), 9 deletions(-)
xen/drivers/char/console.c +9 -9
index d67e1993f2..f49c6f29a8 100644
@@ -346,7 +346,7 @@ void console_giveback(int id)
         serial_steal_fn = NULL;
 }
 
-static void sercon_puts(const char *s, size_t nr)
+void console_serial_puts(const char *s, size_t nr)
 {
     if ( serial_steal_fn != NULL )
         serial_steal_fn(s, nr);
@@ -388,7 +388,7 @@ static void dump_console_ring_key(unsigned char key)
         c += len;
     }
 
-    sercon_puts(buf, sofar);
+    console_serial_puts(buf, sofar);
     video_puts(buf, sofar);
 
     free_xenheap_pages(buf, order);
@@ -547,7 +547,7 @@ static long guest_console_write(XEN_GUEST_HANDLE_PARAM(char) buffer,
             /* Use direct console output as it could be interactive */
             spin_lock_irq(&console_lock);
 
-            sercon_puts(kbuf, kcount);
+            console_serial_puts(kbuf, kcount);
             video_puts(kbuf, kcount);
 
 #ifdef CONFIG_X86
@@ -674,7 +674,7 @@ static void __putstr(const char *str)
 
     ASSERT(spin_is_locked(&console_lock));
 
-    sercon_puts(str, len);
+    console_serial_puts(str, len);
     video_puts(str, len);
 
 #ifdef CONFIG_X86
@@ -1186,12 +1186,12 @@ static void debugtrace_dump_worker(void)
     /* Print oldest portion of the ring. */
     ASSERT(debugtrace_buf[debugtrace_bytes - 1] == 0);
     if ( debugtrace_buf[debugtrace_prd] != '\0' )
-        sercon_puts(&debugtrace_buf[debugtrace_prd],
-                    debugtrace_bytes - debugtrace_prd - 1);
+        console_serial_puts(&debugtrace_buf[debugtrace_prd],
+                            debugtrace_bytes - debugtrace_prd - 1);
 
     /* Print youngest portion of the ring. */
     debugtrace_buf[debugtrace_prd] = '\0';
-    sercon_puts(&debugtrace_buf[0], debugtrace_prd);
+    console_serial_puts(&debugtrace_buf[0], debugtrace_prd);
 
     memset(debugtrace_buf, '\0', debugtrace_bytes);
 
@@ -1274,8 +1274,8 @@ void debugtrace_printk(const char *fmt, ...)
     {
         unsigned int n = scnprintf(cntbuf, sizeof(cntbuf), "%u ", ++count);
 
-        serial_puts(sercon_handle, cntbuf, n);
-        serial_puts(sercon_handle, buf, nr);
+        console_serial_puts(cntbuf, n);
+        console_serial_puts(buf, nr);
     }
     else
     {
xen/include/xen/console.h +3 -0
index ba914f9e5b..53c56191ba 100644
@@ -46,6 +46,9 @@ void console_giveback(int id);
 int console_suspend(void);
 int console_resume(void);
 
+/* Emit a string via the serial console. */
+void console_serial_puts(const char *s, size_t nr);
+
 extern int8_t opt_console_xen;
 
 #endif /* __CONSOLE_H__ */
Signature · 7 lines
2.16.4


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#2 lore raw

Re: [Xen-devel] [PATCH v3 1/4] xen: use common output function in debugtrace

Quoted 6 lines · Juergen Gross

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

No matter that I like this change, I'm not convinced it's correct:
There are two differences between the functions, neither of which I could qualify as "not really making sense": Why is it obvious that it makes no sense for the debugtrace output to bypass one or both of serial_steal_fn() and pv_console_puts()? If you're convinced of this, please provide the "why"-s behind the sentence above.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#3 lore raw

Re: [Xen-devel] [PATCH v3 1/4] xen: use common output function in debugtrace

Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

Quoted 7 lines

No matter that I like this change, I'm not convinced it's correct:
There are two differences between the functions, neither of which I could qualify as "not really making sense": Why is it obvious that it makes no sense for the debugtrace output to bypass one or both of serial_steal_fn() and pv_console_puts()? If you're convinced of this, please provide the "why"-s behind the sentence above.

Okay, I'll use:

Use sercon_puts() in both cases, to be consistent between dumping the buffer when switching debugtrace to console output and when printing a debugtrace entry directly to console.

Juergen

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#4 lore raw

Re: [Xen-devel] [PATCH v3 1/4] xen: use common output function in debugtrace

Quoted 17 lines · Juergen Gross
Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

Quoted 7 lines

No matter that I like this change, I'm not convinced it's correct:
There are two differences between the functions, neither of which I could qualify as "not really making sense": Why is it obvious that it makes no sense for the debugtrace output to bypass one or both of serial_steal_fn() and pv_console_puts()? If you're convinced of this, please provide the "why"-s behind the sentence above.

Quoted 5 lines

Okay, I'll use:

Use sercon_puts() in both cases, to be consistent between dumping the buffer when switching debugtrace to console output and when printing a debugtrace entry directly to console.

Well, this is better as an explanation indeed, but it still doesn't make clear whether it wasn't in fact wanted for there to be a difference in where output gets sent. I may believe that bypassing pv_console_puts() wasn't intended, but the steal function bypass had been there in 3.2 already, so may well have been on purpose.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#5 lore raw

Re: [Xen-devel] [PATCH v3 1/4] xen: use common output function in debugtrace

Quoted 24 lines · Jan Beulich
Quoted 17 lines · Juergen Gross
Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

Quoted 7 lines

No matter that I like this change, I'm not convinced it's correct:
There are two differences between the functions, neither of which I could qualify as "not really making sense": Why is it obvious that it makes no sense for the debugtrace output to bypass one or both of serial_steal_fn() and pv_console_puts()? If you're convinced of this, please provide the "why"-s behind the sentence above.

Quoted 5 lines

Okay, I'll use:

Use sercon_puts() in both cases, to be consistent between dumping the buffer when switching debugtrace to console output and when printing a debugtrace entry directly to console.

Quoted 5 lines

Well, this is better as an explanation indeed, but it still doesn't make clear whether it wasn't in fact wanted for there to be a difference in where output gets sent. I may believe that bypassing pv_console_puts() wasn't intended, but the steal function bypass had been there in 3.2 already, so may well have been on purpose.

There are two users of console_steal():

suspend handling - I believe it is a good idea to not use the serial interface while it is potentially uninitialized.

gdb support - Why should that be special? Not treating the output data appropriate to the attached debugger will be worse than encapsulating it in a way the debugger can handle.

Juergen

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#6 lore raw

Re: [Xen-devel] [PATCH v3 1/4] xen: use common output function in debugtrace

Quoted 31 lines · Juergen Gross
Quoted 24 lines · Jan Beulich
Quoted 17 lines · Juergen Gross
Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

Quoted 7 lines

No matter that I like this change, I'm not convinced it's correct:
There are two differences between the functions, neither of which I could qualify as "not really making sense": Why is it obvious that it makes no sense for the debugtrace output to bypass one or both of serial_steal_fn() and pv_console_puts()? If you're convinced of this, please provide the "why"-s behind the sentence above.

Quoted 5 lines

Okay, I'll use:

Use sercon_puts() in both cases, to be consistent between dumping the buffer when switching debugtrace to console output and when printing a debugtrace entry directly to console.

Quoted 5 lines

Well, this is better as an explanation indeed, but it still doesn't make clear whether it wasn't in fact wanted for there to be a difference in where output gets sent. I may believe that bypassing pv_console_puts() wasn't intended, but the steal function bypass had been there in 3.2 already, so may well have been on purpose.

Quoted 4 lines

There are two users of console_steal():

suspend handling - I believe it is a good idea to not use the serial interface while it is potentially uninitialized.

I agree here.

Quoted 3 lines

gdb support - Why should that be special? Not treating the output data appropriate to the attached debugger will be worse than encapsulating it in a way the debugger can handle.

I'm not sure here - it may well have been intentional to avoid cluttering the debugger console.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#7 lore raw

Re: [Xen-devel] [PATCH v3 1/4] xen: use common output function in debugtrace

Quoted 37 lines · Jan Beulich
Quoted 31 lines · Juergen Gross
Quoted 24 lines · Jan Beulich
Quoted 17 lines · Juergen Gross
Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

Quoted 7 lines

No matter that I like this change, I'm not convinced it's correct:
There are two differences between the functions, neither of which I could qualify as "not really making sense": Why is it obvious that it makes no sense for the debugtrace output to bypass one or both of serial_steal_fn() and pv_console_puts()? If you're convinced of this, please provide the "why"-s behind the sentence above.

Quoted 5 lines

Okay, I'll use:

Use sercon_puts() in both cases, to be consistent between dumping the buffer when switching debugtrace to console output and when printing a debugtrace entry directly to console.

Quoted 5 lines

Well, this is better as an explanation indeed, but it still doesn't make clear whether it wasn't in fact wanted for there to be a difference in where output gets sent. I may believe that bypassing pv_console_puts() wasn't intended, but the steal function bypass had been there in 3.2 already, so may well have been on purpose.

Quoted 4 lines

There are two users of console_steal():

suspend handling - I believe it is a good idea to not use the serial interface while it is potentially uninitialized.

Quoted 8 lines

I agree here.

Quoted 3 lines

gdb support - Why should that be special? Not treating the output data appropriate to the attached debugger will be worse than encapsulating it in a way the debugger can handle.

I'm not sure here - it may well have been intentional to avoid cluttering the debugger console.

But won't using serial_puts() clutter the debugger console? "normal" printk() output seem to be handled in a special way with gdb active, while just using serial_puts() with attached gdb is looking at least suspicious to me. There seems to be a special format, e.g. text output of the hypervisor is prepended with 'O' and the data is sent in hex representation. I can't imagine just sending some ASCII text is the desired approach for debugtrace output.

Juergen

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

Reviewed-by
#8 lore raw

Re: [Xen-devel] [PATCH v3 1/4] xen: use common output function in debugtrace

Quoted 47 lines · Juergen Gross
Quoted 37 lines · Jan Beulich
Quoted 31 lines · Juergen Gross
Quoted 24 lines · Jan Beulich
Quoted 17 lines · Juergen Gross
Quoted 8 lines · Jan Beulich
Quoted 6 lines · Juergen Gross

Today dumping the debugtrace buffers is done via sercon_puts(), while direct printing of trace entries (after toggling output to the console) is using serial_puts().

Use sercon_puts() in both cases, as the difference between both is not really making sense.

Quoted 7 lines

No matter that I like this change, I'm not convinced it's correct:
There are two differences between the functions, neither of which I could qualify as "not really making sense": Why is it obvious that it makes no sense for the debugtrace output to bypass one or both of serial_steal_fn() and pv_console_puts()? If you're convinced of this, please provide the "why"-s behind the sentence above.

Quoted 5 lines

Okay, I'll use:

Use sercon_puts() in both cases, to be consistent between dumping the buffer when switching debugtrace to console output and when printing a debugtrace entry directly to console.

Quoted 5 lines

Well, this is better as an explanation indeed, but it still doesn't make clear whether it wasn't in fact wanted for there to be a difference in where output gets sent. I may believe that bypassing pv_console_puts() wasn't intended, but the steal function bypass had been there in 3.2 already, so may well have been on purpose.

Quoted 4 lines

There are two users of console_steal():

suspend handling - I believe it is a good idea to not use the serial interface while it is potentially uninitialized.

Quoted 8 lines

I agree here.

Quoted 3 lines

gdb support - Why should that be special? Not treating the output data appropriate to the attached debugger will be worse than encapsulating it in a way the debugger can handle.

I'm not sure here - it may well have been intentional to avoid cluttering the debugger console.

Quoted 7 lines

But won't using serial_puts() clutter the debugger console? "normal" printk() output seem to be handled in a special way with gdb active, while just using serial_puts() with attached gdb is looking at least suspicious to me. There seems to be a special format, e.g. text output of the hypervisor is prepended with 'O' and the data is sent in hex representation. I can't imagine just sending some ASCII text is the desired approach for debugtrace output.

Oh, did I get it backwards, I'm sorry. Yes, I agree. With the slightly adjusted description

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#9 lore raw

PATCH v3 2/4 xen: move debugtrace coding to common/debugtrace.c

In reply to Juergen Gross

Instead of living in drivers/char/console.c move the debugtrace related coding to a new file common/debugtrace.c

No functional change, code movement only.

---
 xen/common/Makefile        |   1 +
 xen/common/debugtrace.c    | 181 +++++++++++++++++++++++++++++++++++++++++++++
 xen/drivers/char/console.c | 178 +-------------------------------------------
 3 files changed, 183 insertions(+), 177 deletions(-)
 create mode 100644 xen/common/debugtrace.c
xen/common/Makefile +1 -0
index eddda5daa6..62b34e69e9 100644
@@ -4,6 +4,7 @@ obj-y += bsearch.o
 obj-$(CONFIG_CORE_PARKING) += core_parking.o
 obj-y += cpu.o
 obj-y += cpupool.o
+obj-$(CONFIG_DEBUG_TRACE) += debugtrace.o
 obj-$(CONFIG_HAS_DEVICE_TREE) += device_tree.o
 obj-y += domctl.o
 obj-y += domain.o
xen/common/debugtrace.c new file +181 -0

181 lines not on this page · raw message

xen/drivers/char/console.c +1 -177

184 lines not on this page · raw message

Signature · 7 lines
2.16.4


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Acked-by on v3 2/4
#10 lore raw

Re: [Xen-devel] [PATCH v3 2/4] xen: move debugtrace coding to common/debugtrace.c

Quoted 6 lines · Juergen Gross

Instead of living in drivers/char/console.c move the debugtrace related coding to a new file common/debugtrace.c

No functional change, code movement only.

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#11 lore raw

PATCH v3 3/4 xen: refactor debugtrace data

In reply to Juergen Gross

As a preparation for per-cpu buffers do a little refactoring of the debugtrace data: put the needed buffer admin data into the buffer as it will be needed for each buffer.

While at it switch debugtrace_send_to_console and debugtrace_used to bool and delete an empty line.

---
 xen/common/debugtrace.c | 62 ++++++++++++++++++++++++++++---------------------
 1 file changed, 35 insertions(+), 27 deletions(-)
xen/common/debugtrace.c +35 -27
index a2fab0d72c..7a9e4fb99f 100644
@@ -15,33 +15,39 @@
 #include <xen/watchdog.h>
 
 /* Send output direct to console, or buffer it? */
-static volatile int debugtrace_send_to_console;
+static volatile bool debugtrace_send_to_console;
 
-static char        *debugtrace_buf; /* Debug-trace buffer */
-static unsigned int debugtrace_prd; /* Producer index     */
-static unsigned int debugtrace_kilobytes = 128, debugtrace_bytes;
-static unsigned int debugtrace_used;
+struct debugtrace_data_s {
+    unsigned long bytes; /* Size of buffer. */
+    unsigned long prd;   /* Producer index. */
+    char          buf[];
+};
+
+static struct debugtrace_data_s *debtr_data;
+
+static unsigned int debugtrace_kilobytes = 128;
+static bool debugtrace_used;
 static DEFINE_SPINLOCK(debugtrace_lock);
 integer_param("debugtrace", debugtrace_kilobytes);
 
 static void debugtrace_dump_worker(void)
 {
-    if ( (debugtrace_bytes == 0) || !debugtrace_used )
+    if ( !debtr_data || !debugtrace_used )
         return;
 
     printk("debugtrace_dump() starting\n");
 
     /* Print oldest portion of the ring. */
-    ASSERT(debugtrace_buf[debugtrace_bytes - 1] == 0);
-    if ( debugtrace_buf[debugtrace_prd] != '\0' )
-        console_serial_puts(&debugtrace_buf[debugtrace_prd],
-                            debugtrace_bytes - debugtrace_prd - 1);
+    ASSERT(debtr_data->buf[debtr_data->bytes - 1] == 0);
+    if ( debtr_data->buf[debtr_data->prd] != '\0' )
+        console_serial_puts(&debtr_data->buf[debtr_data->prd],
+                            debtr_data->bytes - debtr_data->prd - 1);
 
     /* Print youngest portion of the ring. */
-    debugtrace_buf[debugtrace_prd] = '\0';
-    console_serial_puts(&debugtrace_buf[0], debugtrace_prd);
+    debtr_data->buf[debtr_data->prd] = '\0';
+    console_serial_puts(&debtr_data->buf[0], debtr_data->prd);
 
-    memset(debugtrace_buf, '\0', debugtrace_bytes);
+    memset(debtr_data->buf, '\0', debtr_data->bytes);
 
     printk("debugtrace_dump() finished\n");
 }
@@ -66,7 +72,6 @@ static void debugtrace_toggle(void)
 
     spin_unlock_irqrestore(&debugtrace_lock, flags);
     watchdog_enable();
-
 }
 
 void debugtrace_dump(void)
@@ -88,10 +93,10 @@ static void debugtrace_add_to_buf(char *buf)
 
     for ( p = buf; *p != '\0'; p++ )
     {
-        debugtrace_buf[debugtrace_prd++] = *p;
+        debtr_data->buf[debtr_data->prd++] = *p;
         /* Always leave a nul byte at the end of the buffer. */
-        if ( debugtrace_prd == (debugtrace_bytes - 1) )
-            debugtrace_prd = 0;
+        if ( debtr_data->prd == (debtr_data->bytes - 1) )
+            debtr_data->prd = 0;
     }
 }
 
@@ -105,14 +110,14 @@ void debugtrace_printk(const char *fmt, ...)
     unsigned long flags;
     unsigned int nr;
 
-    if ( debugtrace_bytes == 0 )
+    if ( !debtr_data )
         return;
 
-    debugtrace_used = 1;
+    debugtrace_used = true;
 
     spin_lock_irqsave(&debugtrace_lock, flags);
 
-    ASSERT(debugtrace_buf[debugtrace_bytes - 1] == 0);
+    ASSERT(debtr_data->buf[debtr_data->bytes - 1] == 0);
 
     va_start(args, fmt);
     nr = vsnprintf(buf, sizeof(buf), fmt, args);
@@ -129,14 +134,14 @@ void debugtrace_printk(const char *fmt, ...)
     {
         if ( strcmp(buf, last_buf) )
         {
-            last_prd = debugtrace_prd;
+            last_prd = debtr_data->prd;
             last_count = ++count;
             safe_strcpy(last_buf, buf);
             snprintf(cntbuf, sizeof(cntbuf), "%u ", count);
         }
         else
         {
-            debugtrace_prd = last_prd;
+            debtr_data->prd = last_prd;
             snprintf(cntbuf, sizeof(cntbuf), "%u-%u ", last_count, ++count);
         }
         debugtrace_add_to_buf(cntbuf);
@@ -154,7 +159,8 @@ static void debugtrace_key(unsigned char key)
 static int __init debugtrace_init(void)
 {
     int order;
-    unsigned int kbytes, bytes;
+    unsigned long kbytes, bytes;
+    struct debugtrace_data_s *data;
 
     /* Round size down to next power of two. */
     while ( (kbytes = (debugtrace_kilobytes & (debugtrace_kilobytes-1))) != 0 )
@@ -165,12 +171,14 @@ static int __init debugtrace_init(void)
         return 0;
 
     order = get_order_from_bytes(bytes);
-    debugtrace_buf = alloc_xenheap_pages(order, 0);
-    ASSERT(debugtrace_buf != NULL);
+    data = alloc_xenheap_pages(order, 0);
+    if ( !data )
+        return -ENOMEM;
 
-    memset(debugtrace_buf, '\0', bytes);
+    memset(data, '\0', bytes);
 
-    debugtrace_bytes = bytes;
+    data->bytes = bytes - sizeof(*data);
+    debtr_data = data;
 
     register_keyhandler('T', debugtrace_key,
                         "toggle debugtrace to console/buffer", 0);
Signature · 7 lines
2.16.4


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#12 lore raw

Re: [Xen-devel] [PATCH v3 3/4] xen: refactor debugtrace data

Quoted 18 lines · Juergen Gross

As a preparation for per-cpu buffers do a little refactoring of the debugtrace data: put the needed buffer admin data into the buffer as it will be needed for each buffer.

While at it switch debugtrace_send_to_console and debugtrace_used to bool and delete an empty line.

---
 xen/common/debugtrace.c | 62 ++++++++++++++++++++++++++++---------------------
 1 file changed, 35 insertions(+), 27 deletions(-)
xen/common/debugtrace.c +0 -0
index a2fab0d72c..7a9e4fb99f 100644
@@ -15,33 +15,39 @@
 #include <xen/watchdog.h>

xen/common/debugtrace.c line 20

Quoted 10 lines
xen/common/debugtrace.c +2 -5
 
 /* Send output direct to console, or buffer it? */
-static volatile int debugtrace_send_to_console;
+static volatile bool debugtrace_send_to_console;
 
-static char        *debugtrace_buf; /* Debug-trace buffer */
-static unsigned int debugtrace_prd; /* Producer index     */
-static unsigned int debugtrace_kilobytes = 128, debugtrace_bytes;
-static unsigned int debugtrace_used;
+struct debugtrace_data_s {

Is the _s suffix really needed for some reason?

xen/common/debugtrace.c line 22

Quoted 2 lines
xen/common/debugtrace.c +2 -0
+    unsigned long bytes; /* Size of buffer. */
+    unsigned long prd;   /* Producer index. */

Seeing that you switch from int to long here - are you really fancying these buffers to go beyond 4GiB in size?

xen/common/debugtrace.c line 26

Quoted 4 lines
xen/common/debugtrace.c +4 -0
+    char          buf[];
+};
+
+static struct debugtrace_data_s *debtr_data;

How about s/debtr/dt/ or s/debtr/debugtrace/ ?

xen/common/debugtrace.c line 28

Quoted 1 line
xen/common/debugtrace.c +1 -0
+static unsigned int debugtrace_kilobytes = 128;

Since you touch this anyway, add __initdata? Maybe also move it next to its integer_param()?

xen/common/debugtrace.c line 36

Quoted 9 lines
xen/common/debugtrace.c +2 -1
+static bool debugtrace_used;
 static DEFINE_SPINLOCK(debugtrace_lock);
 integer_param("debugtrace", debugtrace_kilobytes);
 
 static void debugtrace_dump_worker(void)
 {
-    if ( (debugtrace_bytes == 0) || !debugtrace_used )
+    if ( !debtr_data || !debugtrace_used )
         return;

Why don't you get rid of the left side of the || altogether?

Quoted 6 lines
xen/common/debugtrace.c +0 -2
@@ -165,12 +171,14 @@ static int __init debugtrace_init(void)
         return 0;
 
     order = get_order_from_bytes(bytes);
-    debugtrace_buf = alloc_xenheap_pages(order, 0);
-    ASSERT(debugtrace_buf != NULL);

xen/common/debugtrace.c static int __init debugtrace_init(void) · line 181

Quoted 10 lines
xen/common/debugtrace.c +6 -2
+    data = alloc_xenheap_pages(order, 0);
+    if ( !data )
+        return -ENOMEM;
 
-    memset(debugtrace_buf, '\0', bytes);
+    memset(data, '\0', bytes);
 
-    debugtrace_bytes = bytes;
+    data->bytes = bytes - sizeof(*data);
+    debtr_data = data;

The allocation may end up being almost twice as big as what gets actually used this way. Wouldn't it make sense to re-calculate "bytes" from "order"?

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#13 lore raw

Re: [Xen-devel] [PATCH v3 3/4] xen: refactor debugtrace data

Quoted 24 lines · Jan Beulich
Quoted 23 lines · Juergen Gross

As a preparation for per-cpu buffers do a little refactoring of the debugtrace data: put the needed buffer admin data into the buffer as it will be needed for each buffer.

While at it switch debugtrace_send_to_console and debugtrace_used to bool and delete an empty line.

---
  xen/common/debugtrace.c | 62 ++++++++++++++++++++++++++++---------------------
  1 file changed, 35 insertions(+), 27 deletions(-)
xen/common/debugtrace.c +1 -1
index a2fab0d72c..7a9e4fb99f 100644
@@ -15,33 +15,39 @@
  #include <xen/watchdog.h>
 
  /* Send output direct to console, or buffer it? */
-static volatile int debugtrace_send_to_console;
+static volatile bool debugtrace_send_to_console;
 
Quoted 7 lines

xen/common/debugtrace.c line 20

Quoted 5 lines
xen/common/debugtrace.c +1 -4
-static char        *debugtrace_buf; /* Debug-trace buffer */
-static unsigned int debugtrace_prd; /* Producer index     */
-static unsigned int debugtrace_kilobytes = 128, debugtrace_bytes;
-static unsigned int debugtrace_used;
+struct debugtrace_data_s {

Is the _s suffix really needed for some reason?

No, not really. I can drop it.

Quoted 6 lines

xen/common/debugtrace.c line 22

Quoted 3 lines
xen/common/debugtrace.c +2 -0
+    unsigned long bytes; /* Size of buffer. */
+    unsigned long prd;   /* Producer index. */

Seeing that you switch from int to long here - are you really fancying these buffers to go beyond 4GiB in size?

I didn't want to rule it out on a multi-TB machine. :-)

Quoted 7 lines

xen/common/debugtrace.c line 26

Quoted 5 lines
xen/common/debugtrace.c +4 -0
+    char          buf[];
+};
+
+static struct debugtrace_data_s *debtr_data;

How about s/debtr/dt/ or s/debtr/debugtrace/ ?

Yes, dt seems to be fine.

Quoted 5 lines

xen/common/debugtrace.c line 28

Quoted 2 lines
xen/common/debugtrace.c +1 -0
+static unsigned int debugtrace_kilobytes = 128;

Since you touch this anyway, add __initdata? Maybe also move it next to its integer_param()?

This is modified in patch 4 again, and there I need it in the running system for support of cpu hotplug with per-cpu buffers.

Quoted 5 lines
Quoted 5 lines
xen/common/debugtrace.c +1 -0
+static bool debugtrace_used;
  static DEFINE_SPINLOCK(debugtrace_lock);
  integer_param("debugtrace", debugtrace_kilobytes);
 
Quoted 7 lines

xen/common/debugtrace.c line 36

Quoted 5 lines
xen/common/debugtrace.c +1 -1
  static void debugtrace_dump_worker(void)
  {
-    if ( (debugtrace_bytes == 0) || !debugtrace_used )
+    if ( !debtr_data || !debugtrace_used )
          return;

Why don't you get rid of the left side of the || altogether?

In patch 4 I do. :-)

I can drop it here already.

Quoted 14 lines
Quoted 14 lines
xen/common/debugtrace.c +4 -3
@@ -165,12 +171,14 @@ static int __init debugtrace_init(void)
          return 0;
 
      order = get_order_from_bytes(bytes);
-    debugtrace_buf = alloc_xenheap_pages(order, 0);
-    ASSERT(debugtrace_buf != NULL);
+    data = alloc_xenheap_pages(order, 0);
+    if ( !data )
+        return -ENOMEM;
 
-    memset(debugtrace_buf, '\0', bytes);
+    memset(data, '\0', bytes);
 
Quoted 7 lines

xen/common/debugtrace.c static int __init debugtrace_init(void) · line 181

Quoted 3 lines
xen/common/debugtrace.c +2 -1
-    debugtrace_bytes = bytes;
+    data->bytes = bytes - sizeof(*data);
+    debtr_data = data;

The allocation may end up being almost twice as big as what gets actually used this way. Wouldn't it make sense to re-calculate "bytes" from "order"?

Yes, you are right.

Juergen

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#14 lore raw

Re: [Xen-devel] [PATCH v3 3/4] xen: refactor debugtrace data

Quoted 9 lines · Juergen Gross
Quoted 5 lines · Jan Beulich
Quoted 1 line · Juergen Gross

+static unsigned int debugtrace_kilobytes = 128;

Since you touch this anyway, add __initdata? Maybe also move it next to its integer_param()?

This is modified in patch 4 again, and there I need it in the running system for support of cpu hotplug with per-cpu buffers.

Right, I've meanwhile noticed this. Hence it's fine to keep as is.

Quoted 21 lines
Quoted 13 lines
Quoted 13 lines
xen/common/debugtrace.c +4 -3
@@ -165,12 +171,14 @@ static int __init debugtrace_init(void)
          return 0;
 
      order = get_order_from_bytes(bytes);
-    debugtrace_buf = alloc_xenheap_pages(order, 0);
-    ASSERT(debugtrace_buf != NULL);
+    data = alloc_xenheap_pages(order, 0);
+    if ( !data )
+        return -ENOMEM;
 
-    memset(debugtrace_buf, '\0', bytes);
+    memset(data, '\0', bytes);
 
Quoted 7 lines

xen/common/debugtrace.c static int __init debugtrace_init(void) · line 181

Quoted 3 lines
xen/common/debugtrace.c +2 -1
-    debugtrace_bytes = bytes;
+    data->bytes = bytes - sizeof(*data);
+    debtr_data = data;

The allocation may end up being almost twice as big as what gets actually used this way. Wouldn't it make sense to re-calculate "bytes" from "order"?

Quoted 1 line

Yes, you are right.

Actually I wasn't, which I did notice seeing the relevant piece of code getting touched in patch 4:

    while ( (kbytes = (debugtrace_kilobytes & (debugtrace_kilobytes-1))) != 0 )
        debugtrace_kilobytes = kbytes;

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#15 lore raw

Re: [Xen-devel] [PATCH v3 3/4] xen: refactor debugtrace data

Quoted 34 lines · Jan Beulich
Quoted 9 lines · Juergen Gross
Quoted 5 lines · Jan Beulich
Quoted 1 line · Juergen Gross

+static unsigned int debugtrace_kilobytes = 128;

Since you touch this anyway, add __initdata? Maybe also move it next to its integer_param()?

This is modified in patch 4 again, and there I need it in the running system for support of cpu hotplug with per-cpu buffers.

Right, I've meanwhile noticed this. Hence it's fine to keep as is.

Quoted 21 lines
Quoted 20 lines
Quoted 6 lines
xen/common/debugtrace.c +0 -2
@@ -165,12 +171,14 @@ static int __init debugtrace_init(void)
           return 0;
 
       order = get_order_from_bytes(bytes);
-    debugtrace_buf = alloc_xenheap_pages(order, 0);
-    ASSERT(debugtrace_buf != NULL);

xen/common/debugtrace.c static int __init debugtrace_init(void) · line 181

Quoted 10 lines
xen/common/debugtrace.c +6 -2
+    data = alloc_xenheap_pages(order, 0);
+    if ( !data )
+        return -ENOMEM;
 
-    memset(debugtrace_buf, '\0', bytes);
+    memset(data, '\0', bytes);
 
-    debugtrace_bytes = bytes;
+    data->bytes = bytes - sizeof(*data);
+    debtr_data = data;

The allocation may end up being almost twice as big as what gets actually used this way. Wouldn't it make sense to re-calculate "bytes" from "order"?

Quoted 7 lines
Quoted 1 line

Yes, you are right.

Actually I wasn't, which I did notice seeing the relevant piece of code getting touched in patch 4:

     while ( (kbytes = (debugtrace_kilobytes & (debugtrace_kilobytes-1))) != 0 )
         debugtrace_kilobytes = kbytes;

For kbytes < 4 you still were right.

Juergen

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#16 lore raw

PATCH v3 4/4 xen: add per-cpu buffer option to debugtrace

In reply to Juergen Gross

debugtrace is normally writing trace entries into a single trace buffer. There are cases where this is not optimal, e.g. when hunting a bug which requires writing lots of trace entries and one cpu is stuck. This will result in other cpus filling the trace buffer and finally overwriting the interesting trace entries of the hanging cpu.

In order to be able to debug such situations add the capability to use per-cpu trace buffers. This can be selected by specifying the debugtrace boot parameter with the modifier "cpu:", like:

debugtrace=cpu:16

At the same time switch the parsing function to accept size modifiers (e.g. 4M or 1G).

Printing out the trace entries is done for each buffer in order to minimize the effort needed during printing. As each entry is prefixed with its sequence number sorting the entries can easily be done when analyzing them.

---
V2: only allocate buffer if not already done so

---
 docs/misc/xen-command-line.pandoc |   7 +-
 xen/common/debugtrace.c           | 148 ++++++++++++++++++++++++++++----------
 2 files changed, 116 insertions(+), 39 deletions(-)
docs/misc/xen-command-line.pandoc +4 -3
index 7c72e31032..832797e2e2 100644
@@ -644,12 +644,13 @@ over the PCI busses sequentially) or by PCI device (must be on segment 0).
 Limits the number lines printed in Xen stack traces.
 
 ### debugtrace
-> `= <integer>`
+> `= [cpu:]<size>`
 
 > Default: `128`
 
-Specify the size of the console debug trace buffer in KiB. The debug
-trace feature is only enabled in debugging builds of Xen.
+Specify the size of the console debug trace buffer. By specifying `cpu:`
+additionally a trace buffer of the specified size is allocated per cpu.
+The debug trace feature is only enabled in debugging builds of Xen.
 
 ### dma_bits
 > `= <integer>`
xen/common/debugtrace.c +112 -36

234 lines not on this page · raw message

Signature · 7 lines
2.16.4


_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
#17 lore raw

Re: [Xen-devel] [PATCH v3 4/4] xen: add per-cpu buffer option to debugtrace

xen/common/debugtrace.c struct debugtrace_data_s { · line 30

Quoted 8 lines · Juergen Gross
xen/common/debugtrace.c +2 -1
@@ -24,32 +25,62 @@ struct debugtrace_data_s {
 };
 
 static struct debugtrace_data_s *debtr_data;
+static DEFINE_PER_CPU(struct debugtrace_data_s *, debtr_cpu_data);
 
-static unsigned int debugtrace_kilobytes = 128;
+static unsigned int debugtrace_bytes = 128 << 10;

And after patch 3 this is now left as "unsigned int"?

Quoted 4 lines
xen/common/debugtrace.c +1 -1
+static bool debugtrace_per_cpu;
 static bool debugtrace_used;
 static DEFINE_SPINLOCK(debugtrace_lock);
-integer_param("debugtrace", debugtrace_kilobytes);

xen/common/debugtrace.c struct debugtrace_data_s { · line 42

Quoted 10 lines
xen/common/debugtrace.c +8 -1
 
-static void debugtrace_dump_worker(void)
+static int __init debugtrace_parse_param(const char *s)
+{
+    if ( !strncmp(s, "cpu:", 4) )
+    {
+        debugtrace_per_cpu = true;
+        s += 4;
+    }
+    debugtrace_bytes =  parse_size_and_unit(s, NULL);

Stray double blank.

Quoted 12 lines
xen/common/debugtrace.c +7 -2
+    return 0;
+}
+custom_param("debugtrace", debugtrace_parse_param);
+
+static void debugtrace_dump_buffer(struct debugtrace_data_s *data,
+                                   const char *which)
 {
-    if ( !debtr_data || !debugtrace_used )
+    if ( !data )
         return;
 
-    printk("debugtrace_dump() starting\n");

xen/common/debugtrace.c struct debugtrace_data_s { · line 58

Quoted 10 lines
xen/common/debugtrace.c +4 -4
+    printk("debugtrace_dump() %s buffer starting\n", which);
 
     /* Print oldest portion of the ring. */
-    ASSERT(debtr_data->buf[debtr_data->bytes - 1] == 0);
-    if ( debtr_data->buf[debtr_data->prd] != '\0' )
-        console_serial_puts(&debtr_data->buf[debtr_data->prd],
-                            debtr_data->bytes - debtr_data->prd - 1);
+    ASSERT(data->buf[data->bytes - 1] == 0);
+    if ( data->buf[data->prd] != '\0' )
+        console_serial_puts(&data->buf[data->prd], data->bytes - data->prd - 1);

Seeing this getting changed yet another time I now really wonder if this nul termination is really still needed now that a size is being passed into the actual output function. If you got rid of this in an early prereq patch, the subsequent (to that one) ones would shrink.

Furthermore I can't help thinking that said change to pass the size into the actual output functions actually broke the logic here: By memset()-ing the buffer to zero, output on a subsequent invocation would have been suitably truncated (in fact, until prd had wrapped, I think it would have got truncated more than intended). Julien, can you please look into this apparent regression?

Quoted 9 lines
xen/common/debugtrace.c +1 -2
@@ -156,33 +195,70 @@ static void debugtrace_key(unsigned char key)
     debugtrace_toggle();
 }
 
-static int __init debugtrace_init(void)
+static void debugtrace_alloc_buffer(struct debugtrace_data_s **ptr)
 {
     int order;
-    unsigned long kbytes, bytes;

xen/common/debugtrace.c static void debugtrace_key(unsigned char key) · line 203

Quoted 10 lines
xen/common/debugtrace.c +1 -7
     struct debugtrace_data_s *data;
 
-    /* Round size down to next power of two. */
-    while ( (kbytes = (debugtrace_kilobytes & (debugtrace_kilobytes-1))) != 0 )
-        debugtrace_kilobytes = kbytes;
-
-    bytes = debugtrace_kilobytes << 10;
-    if ( bytes == 0 )
-        return 0;
+    if ( debugtrace_bytes < PAGE_SIZE || *ptr )

Why the check against PAGE_SIZE?

xen/common/debugtrace.c static void debugtrace_key(unsigned char key) · line 210

Quoted 9 lines
xen/common/debugtrace.c +4 -2
+        return;
 
-    order = get_order_from_bytes(bytes);
+    order = get_order_from_bytes(debugtrace_bytes);
     data = alloc_xenheap_pages(order, 0);
     if ( !data )
-        return -ENOMEM;
+    {
+        printk("failed to allocate debugtrace buffer\n");

Perhaps better to also indicate which/whose buffer?

Quoted 12 lines
xen/common/debugtrace.c +12 -0
+        return;
+    }
+
+    memset(data, '\0', debugtrace_bytes);
+    data->bytes = debugtrace_bytes - sizeof(*data);
+
+    *ptr = data;
+}
+
+static int debugtrace_cpu_callback(struct notifier_block *nfb,
+                                   unsigned long action, void *hcpu)
+{

xen/common/debugtrace.c static void debugtrace_key(unsigned char key) · line 232

Quoted 10 lines
xen/common/debugtrace.c +10 -0
+    unsigned int cpu = (unsigned long)hcpu;
+
+    /* Buffers are only ever allocated, never freed. */
+    switch ( action )
+    {
+    case CPU_UP_PREPARE:
+        debugtrace_alloc_buffer(&per_cpu(debtr_cpu_data, cpu));
+        break;
+    default:
+        break;

There no apparent need for "default:" here; quite possibly this could be if() instead of switch() anyway.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#18 lore raw

Re: [Xen-devel] [PATCH v3 4/4] xen: add per-cpu buffer option to debugtrace

Quoted 11 lines · Jan Beulich

xen/common/debugtrace.c struct debugtrace_data_s { · line 30

Quoted 8 lines · Juergen Gross
xen/common/debugtrace.c +2 -1
@@ -24,32 +25,62 @@ struct debugtrace_data_s {
  };
 
  static struct debugtrace_data_s *debtr_data;
+static DEFINE_PER_CPU(struct debugtrace_data_s *, debtr_cpu_data);
 
-static unsigned int debugtrace_kilobytes = 128;
+static unsigned int debugtrace_bytes = 128 << 10;

And after patch 3 this is now left as "unsigned int"?

Good catch. :-)

Quoted 6 lines
Quoted 6 lines
xen/common/debugtrace.c +1 -1
+static bool debugtrace_per_cpu;
  static bool debugtrace_used;
  static DEFINE_SPINLOCK(debugtrace_lock);
-integer_param("debugtrace", debugtrace_kilobytes);
 
Quoted 11 lines

xen/common/debugtrace.c struct debugtrace_data_s { · line 42

Quoted 9 lines
xen/common/debugtrace.c +8 -1
-static void debugtrace_dump_worker(void)
+static int __init debugtrace_parse_param(const char *s)
+{
+    if ( !strncmp(s, "cpu:", 4) )
+    {
+        debugtrace_per_cpu = true;
+        s += 4;
+    }
+    debugtrace_bytes =  parse_size_and_unit(s, NULL);

Stray double blank.

Yes. Will remove one.

Quoted 15 lines
Quoted 15 lines
xen/common/debugtrace.c +8 -2
+    return 0;
+}
+custom_param("debugtrace", debugtrace_parse_param);
+
+static void debugtrace_dump_buffer(struct debugtrace_data_s *data,
+                                   const char *which)
  {
-    if ( !debtr_data || !debugtrace_used )
+    if ( !data )
          return;
 
-    printk("debugtrace_dump() starting\n");
+    printk("debugtrace_dump() %s buffer starting\n", which);
 
Quoted 13 lines

xen/common/debugtrace.c struct debugtrace_data_s { · line 58

Quoted 8 lines
xen/common/debugtrace.c +3 -4
      /* Print oldest portion of the ring. */
-    ASSERT(debtr_data->buf[debtr_data->bytes - 1] == 0);
-    if ( debtr_data->buf[debtr_data->prd] != '\0' )
-        console_serial_puts(&debtr_data->buf[debtr_data->prd],
-                            debtr_data->bytes - debtr_data->prd - 1);
+    ASSERT(data->buf[data->bytes - 1] == 0);
+    if ( data->buf[data->prd] != '\0' )
+        console_serial_puts(&data->buf[data->prd], data->bytes - data->prd - 1);

Seeing this getting changed yet another time I now really wonder if this nul termination is really still needed now that a size is being passed into the actual output function. If you got rid of this in an early prereq patch, the subsequent (to that one) ones would shrink.

Yes.

Quoted 7 lines

Furthermore I can't help thinking that said change to pass the size into the actual output functions actually broke the logic here: By memset()-ing the buffer to zero, output on a subsequent invocation would have been suitably truncated (in fact, until prd had wrapped, I think it would have got truncated more than intended). Julien, can you please look into this apparent regression?

I can do that. Resetting prd to 0 when clearing the buffer is required here.

Quoted 12 lines
Quoted 12 lines
xen/common/debugtrace.c +1 -2
@@ -156,33 +195,70 @@ static void debugtrace_key(unsigned char key)
      debugtrace_toggle();
  }
 
-static int __init debugtrace_init(void)
+static void debugtrace_alloc_buffer(struct debugtrace_data_s **ptr)
  {
      int order;
-    unsigned long kbytes, bytes;
      struct debugtrace_data_s *data;
 
Quoted 10 lines

xen/common/debugtrace.c static void debugtrace_key(unsigned char key) · line 203

Quoted 8 lines
xen/common/debugtrace.c +1 -7
-    /* Round size down to next power of two. */
-    while ( (kbytes = (debugtrace_kilobytes & (debugtrace_kilobytes-1))) != 0 )
-        debugtrace_kilobytes = kbytes;
-
-    bytes = debugtrace_kilobytes << 10;
-    if ( bytes == 0 )
-        return 0;
+    if ( debugtrace_bytes < PAGE_SIZE || *ptr )

Why the check against PAGE_SIZE?

With recaclulating debugtrace_bytes this can be dropped.

Quoted 11 lines

xen/common/debugtrace.c static void debugtrace_key(unsigned char key) · line 210

Quoted 10 lines
xen/common/debugtrace.c +4 -2
+        return;
 
-    order = get_order_from_bytes(bytes);
+    order = get_order_from_bytes(debugtrace_bytes);
      data = alloc_xenheap_pages(order, 0);
      if ( !data )
-        return -ENOMEM;
+    {
+        printk("failed to allocate debugtrace buffer\n");
Quoted 1 line

Perhaps better to also indicate which/whose buffer?

Hmm, I'm not sure this is really required. I can add it if you want, but as a user of debugtrace I'd be only interested in the information whether I can expect all trace entries to be seen or not.

Quoted 18 lines
Quoted 18 lines
xen/common/debugtrace.c +17 -0
+        return;
+    }
+
+    memset(data, '\0', debugtrace_bytes);
+    data->bytes = debugtrace_bytes - sizeof(*data);
+
+    *ptr = data;
+}
+
+static int debugtrace_cpu_callback(struct notifier_block *nfb,
+                                   unsigned long action, void *hcpu)
+{
+    unsigned int cpu = (unsigned long)hcpu;
+
+    /* Buffers are only ever allocated, never freed. */
+    switch ( action )
+    {
Quoted 8 lines

xen/common/debugtrace.c static void debugtrace_key(unsigned char key) · line 232

Quoted 5 lines
xen/common/debugtrace.c +5 -0
+    case CPU_UP_PREPARE:
+        debugtrace_alloc_buffer(&per_cpu(debtr_cpu_data, cpu));
+        break;
+    default:
+        break;

There no apparent need for "default:" here; quite possibly this could be if() instead of switch() anyway.

Fine with me.

Juergen

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#19 lore raw

Re: [Xen-devel] [PATCH v3 4/4] xen: add per-cpu buffer option to debugtrace

Quoted 36 lines · Juergen Gross
Quoted 20 lines · Jan Beulich
Quoted 8 lines · Juergen Gross
xen/common/debugtrace.c +3 -2
+static void debugtrace_dump_buffer(struct debugtrace_data_s *data,
+                                   const char *which)
  {
-    if ( !debtr_data || !debugtrace_used )
+    if ( !data )
          return;
 
-    printk("debugtrace_dump() starting\n");

xen/common/debugtrace.c struct debugtrace_data_s { · line 58

Quoted 10 lines
xen/common/debugtrace.c +4 -4
+    printk("debugtrace_dump() %s buffer starting\n", which);
 
      /* Print oldest portion of the ring. */
-    ASSERT(debtr_data->buf[debtr_data->bytes - 1] == 0);
-    if ( debtr_data->buf[debtr_data->prd] != '\0' )
-        console_serial_puts(&debtr_data->buf[debtr_data->prd],
-                            debtr_data->bytes - debtr_data->prd - 1);
+    ASSERT(data->buf[data->bytes - 1] == 0);
+    if ( data->buf[data->prd] != '\0' )
+        console_serial_puts(&data->buf[data->prd], data->bytes - data->prd - 1);
Quoted 4 lines

Seeing this getting changed yet another time I now really wonder if this nul termination is really still needed now that a size is being passed into the actual output function. If you got rid of this in an early prereq patch, the subsequent (to that one) ones would shrink.

Yes.

Quoted 7 lines

Furthermore I can't help thinking that said change to pass the size into the actual output functions actually broke the logic here: By memset()-ing the buffer to zero, output on a subsequent invocation would have been suitably truncated (in fact, until prd had wrapped, I think it would have got truncated more than intended). Julien, can you please look into this apparent regression?

Quoted 2 lines

I can do that. Resetting prd to 0 when clearing the buffer is required here.

I'm afraid it's not this simple: Doing so will confuse debugtrace_printk() - consider the function then storing the previously latched last_prd into debugtrace_prd.

Quoted 8 lines
Quoted 8 lines
Quoted 7 lines
xen/common/debugtrace.c +3 -2
-    order = get_order_from_bytes(bytes);
+    order = get_order_from_bytes(debugtrace_bytes);
      data = alloc_xenheap_pages(order, 0);
      if ( !data )
-        return -ENOMEM;
+    {
+        printk("failed to allocate debugtrace buffer\n");
Quoted 5 lines
Quoted 1 line

Perhaps better to also indicate which/whose buffer?

Hmm, I'm not sure this is really required. I can add it if you want, but as a user of debugtrace I'd be only interested in the information whether I can expect all trace entries to be seen or not.

Well, if the allocation fails for a CPU, it's not impossible for the CPU bringup to then also fail. Subsequent to this the system would then still provide an almost complete set of debugtrace entries (ones issued by subsequent bringup actions would be missing), _despite_ this log message.

Jan

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel

#20 lore raw

Re: [Xen-devel] [PATCH v3 4/4] xen: add per-cpu buffer option to debugtrace

Quoted 37 lines · Jan Beulich
Quoted 36 lines · Juergen Gross
Quoted 20 lines · Jan Beulich
Quoted 8 lines · Juergen Gross
xen/common/debugtrace.c +3 -2
+static void debugtrace_dump_buffer(struct debugtrace_data_s *data,
+                                   const char *which)
   {
-    if ( !debtr_data || !debugtrace_used )
+    if ( !data )
           return;
 
-    printk("debugtrace_dump() starting\n");

xen/common/debugtrace.c struct debugtrace_data_s { · line 58

Quoted 10 lines
xen/common/debugtrace.c +4 -4
+    printk("debugtrace_dump() %s buffer starting\n", which);
 
       /* Print oldest portion of the ring. */
-    ASSERT(debtr_data->buf[debtr_data->bytes - 1] == 0);
-    if ( debtr_data->buf[debtr_data->prd] != '\0' )
-        console_serial_puts(&debtr_data->buf[debtr_data->prd],
-                            debtr_data->bytes - debtr_data->prd - 1);
+    ASSERT(data->buf[data->bytes - 1] == 0);
+    if ( data->buf[data->prd] != '\0' )
+        console_serial_puts(&data->buf[data->prd], data->bytes - data->prd - 1);
Quoted 4 lines

Seeing this getting changed yet another time I now really wonder if this nul termination is really still needed now that a size is being passed into the actual output function. If you got rid of this in an early prereq patch, the subsequent (to that one) ones would shrink.

Yes.

Quoted 7 lines

Furthermore I can't help thinking that said change to pass the size into the actual output functions actually broke the logic here: By memset()-ing the buffer to zero, output on a subsequent invocation would have been suitably truncated (in fact, until prd had wrapped, I think it would have got truncated more than intended). Julien, can you please look into this apparent regression?

Quoted 6 lines
Quoted 2 lines

I can do that. Resetting prd to 0 when clearing the buffer is required here.

I'm afraid it's not this simple: Doing so will confuse debugtrace_printk() - consider the function then storing the previously latched last_prd into debugtrace_prd.

Sure, this has to be handled (and it is still easy to do so).

Quoted 15 lines
Quoted 9 lines
Quoted 9 lines
Quoted 8 lines
xen/common/debugtrace.c +3 -2
-    order = get_order_from_bytes(bytes);
+    order = get_order_from_bytes(debugtrace_bytes);
       data = alloc_xenheap_pages(order, 0);
       if ( !data )
-        return -ENOMEM;
+    {
+        printk("failed to allocate debugtrace buffer\n");
Quoted 5 lines
Quoted 1 line

Perhaps better to also indicate which/whose buffer?

Hmm, I'm not sure this is really required. I can add it if you want, but as a user of debugtrace I'd be only interested in the information whether I can expect all trace entries to be seen or not.

Quoted 5 lines

Well, if the allocation fails for a CPU, it's not impossible for the CPU bringup to then also fail. Subsequent to this the system would then still provide an almost complete set of debugtrace entries (ones issued by subsequent bringup actions would be missing), _despite_ this log message.

You seem to want the information, so I can add it.

Juergen

_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel