Skip to contentSkip to conversation

tracing: Ignore mmiotrace from kernel commandline

PATCH In review Ziqian SUN (Zamir) 2 participants 6 messages Sep 11, 2017 – Sep 19, 2017

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v3 1/1 tracing: Ignore mmiotrace from kernel commandline

From: "Ziqian SUN (Zamir)" <[email protected]>

The mmiotrace tracer cannot be enabled with ftrace=mmiotrace in kernel commandline. With this patch, noboot is added to the tracer struct, and when system boot with a tracer that has noboot=true, it will print out a warning message and continue booting.

--
v1 -> v2 : remove unessential comment
v2 -> v3 : Use tracer struct instead of a separate list to store noboot

---
 kernel/trace/trace.c           | 7 +++++++
 kernel/trace/trace.h           | 2 ++
 kernel/trace/trace_mmiotrace.c | 1 +
 3 files changed, 10 insertions(+)
kernel/trace/trace.c +7 -0
index 5360b7a..48c474f 100644
@@ -5358,6 +5358,13 @@ static int tracing_set_tracer(struct trace_array *tr, const char *buf)
 	if (t == tr->current_trace)
 		goto out;
 
+	/* Some tracers won't work on kernel command line */
+	if (system_state < SYSTEM_RUNNING && t->noboot) {
+		pr_warn("Tracer '%s' is not allowed on command line, ignored\n",
+			t->name);
+		goto out;
+	}
+
 	/* Some tracers are only allowed for the top level buffer */
 	if (!trace_ok_for_array(t, tr)) {
 		ret = -EINVAL;
kernel/trace/trace.h +2 -0
index fb5d54d..652c682 100644
@@ -444,6 +444,8 @@ struct tracer {
 #ifdef CONFIG_TRACER_MAX_TRACE
 	bool			use_max_tr;
 #endif
+	/* True if tracer cannot be enabled in kernel param */
+	bool			noboot;
 };
 
 
kernel/trace/trace_mmiotrace.c +1 -0
index cd7480d..dca78fc 100644
@@ -282,6 +282,7 @@ static enum print_line_t mmio_print_line(struct trace_iterator *iter)
 	.close		= mmio_close,
 	.read		= mmio_read,
 	.print_line	= mmio_print_line,
+	.noboot		= true,
 };
 
 __init static int init_mmio_trace(void)
Signature · 1 line
1.8.3.1
#1 lore raw
Quoted 22 lines · Ziqian SUN (Zamir)

From: "Ziqian SUN (Zamir)" <[email protected]>

The mmiotrace tracer cannot be enabled with ftrace=mmiotrace in kernel commandline. With this patch, noboot is added to the tracer struct, and when system boot with a tracer that has noboot=true, it will print out a warning message and continue booting.

--
v1 -> v2 : remove unessential comment
v2 -> v3 : Use tracer struct instead of a separate list to store noboot

---
  kernel/trace/trace.c           | 7 +++++++
  kernel/trace/trace.h           | 2 ++
  kernel/trace/trace_mmiotrace.c | 1 +
  3 files changed, 10 insertions(+)
kernel/trace/trace.c +0 -0
index 5360b7a..48c474f 100644

kernel/trace/trace.c static int tracing_set_tracer(struct trace_array *tr, const char *buf) · line 5364

Quoted 7 lines
kernel/trace/trace.c +4 -0
@@ -5358,6 +5358,13 @@ static int tracing_set_tracer(struct trace_array *tr, const char *buf)
  	if (t == tr->current_trace)
  		goto out;
 
+	/* Some tracers won't work on kernel command line */
+	if (system_state < SYSTEM_RUNNING && t->noboot) {
+		pr_warn("Tracer '%s' is not allowed on command line, ignored\n",
+			t->name);

I feel the core trace printed by WARN is not so meaningful, so I use pr_warn instead.

Quoted 25 lines
kernel/trace/trace.c +3 -0
+		goto out;
+	}
+
  	/* Some tracers are only allowed for the top level buffer */
  	if (!trace_ok_for_array(t, tr)) {
  		ret = -EINVAL;
kernel/trace/trace.h +2 -0
index fb5d54d..652c682 100644
@@ -444,6 +444,8 @@ struct tracer {
  #ifdef CONFIG_TRACER_MAX_TRACE
  	bool			use_max_tr;
  #endif
+	/* True if tracer cannot be enabled in kernel param */
+	bool			noboot;
  };
 
 
kernel/trace/trace_mmiotrace.c +0 -0
index cd7480d..dca78fc 100644

kernel/trace/trace_mmiotrace.c static enum print_line_t mmio_print_line(struct trace_iterator *iter) · line 288

Quoted 7 lines
kernel/trace/trace_mmiotrace.c +1 -0
@@ -282,6 +282,7 @@ static enum print_line_t mmio_print_line(struct trace_iterator *iter)
  	.close		= mmio_close,
  	.read		= mmio_read,
  	.print_line	= mmio_print_line,
+	.noboot		= true,
  };
 
  __init static int init_mmio_trace(void)

This patch passed my own test on top of Fedora kernel 4.13.0-1.fc26.x86_64.

Signature · 6 lines
Ziqian SUN (Zamir)
9F Raycom office (NAY)
Red Hat Software (Beijing) Co.,Ltd
IRC: zsun (internal and freenode)
Tel: +86 10 65627458
GPG : 1D86 6D4A 49CE 4BBD 72CF FCF5 D856 6E11 F2A0 525E
maintainer
#2 lore raw
Quoted 26 lines · On Mon, 11 Sep 2017 14:31:34 +0800 "Ziqian SUN (Zamir)" <[email protected]> wrote:
Quoted 25 lines · Ziqian SUN (Zamir)

From: "Ziqian SUN (Zamir)" <[email protected]>

The mmiotrace tracer cannot be enabled with ftrace=mmiotrace in kernel commandline. With this patch, noboot is added to the tracer struct, and when system boot with a tracer that has noboot=true, it will print out a warning message and continue booting.

--
v1 -> v2 : remove unessential comment
v2 -> v3 : Use tracer struct instead of a separate list to store noboot

---
  kernel/trace/trace.c           | 7 +++++++
  kernel/trace/trace.h           | 2 ++
  kernel/trace/trace_mmiotrace.c | 1 +
  3 files changed, 10 insertions(+)
kernel/trace/trace.c +0 -0
index 5360b7a..48c474f 100644
@@ -5358,6 +5358,13 @@ static int tracing_set_tracer(struct trace_array *tr, const char *buf)
  	if (t == tr->current_trace)
  		goto out;
 
Quoted 6 lines

kernel/trace/trace.c static int tracing_set_tracer(struct trace_array *tr, const char *buf) · line 5364

Quoted 4 lines
kernel/trace/trace.c +4 -0
+	/* Some tracers won't work on kernel command line */
+	if (system_state < SYSTEM_RUNNING && t->noboot) {
+		pr_warn("Tracer '%s' is not allowed on command line, ignored\n",
+			t->name);  

I feel the core trace printed by WARN is not so meaningful, so I use pr_warn instead.

I'm fine with pr_warn, but I'm curious to what you mean by "not so meaningful"? A WARN() will cause a dump stack, which usually shows up as a bug in systems and more likely to be seen. But if someone is adding this to the kernel command line and it's not working, they should be looking for the tracer name within the dmesg anyway.

Also, I'm currently at OSS in LA and hopefully I don't lose this patch. If you don't see anything by next Monday from me, feel free to send me a reminder ping.

-- Steve

#3 lore raw
Quoted 36 lines · Steven Rostedt
Quoted 26 lines · On Mon, 11 Sep 2017 14:31:34 +0800 "Ziqian SUN (Zamir)" <[email protected]> wrote:
Quoted 25 lines · Ziqian SUN (Zamir)

From: "Ziqian SUN (Zamir)" <[email protected]>

The mmiotrace tracer cannot be enabled with ftrace=mmiotrace in kernel commandline. With this patch, noboot is added to the tracer struct, and when system boot with a tracer that has noboot=true, it will print out a warning message and continue booting.

--
v1 -> v2 : remove unessential comment
v2 -> v3 : Use tracer struct instead of a separate list to store noboot

---
   kernel/trace/trace.c           | 7 +++++++
   kernel/trace/trace.h           | 2 ++
   kernel/trace/trace_mmiotrace.c | 1 +
   3 files changed, 10 insertions(+)
kernel/trace/trace.c +0 -0
index 5360b7a..48c474f 100644
@@ -5358,6 +5358,13 @@ static int tracing_set_tracer(struct trace_array *tr, const char *buf)
   	if (t == tr->current_trace)
   		goto out;
 
Quoted 6 lines

kernel/trace/trace.c static int tracing_set_tracer(struct trace_array *tr, const char *buf) · line 5364

Quoted 4 lines
kernel/trace/trace.c +4 -0
+	/* Some tracers won't work on kernel command line */
+	if (system_state < SYSTEM_RUNNING && t->noboot) {
+		pr_warn("Tracer '%s' is not allowed on command line, ignored\n",
+			t->name);

I feel the core trace printed by WARN is not so meaningful, so I use pr_warn instead.

Quoted 6 lines

I'm fine with pr_warn, but I'm curious to what you mean by "not so meaningful"? A WARN() will cause a dump stack, which usually shows up as a bug in systems and more likely to be seen. But if someone is adding this to the kernel command line and it's not working, they should be looking for the tracer name within the dmesg anyway.

Following is the log I tried with WARN before. With WARN we have both the warning message, and the stack trace. The stack trace goes into stacing_set_tracer, but not showing mmiotrace. I mean, this is useful for debugging, but maybe confusing for users who is willing to report bugs but not a kernel geek. With pr_warn we already indicate mmiotrace is ignored, and is neat.

Just my cents.

[    7.134149] Starting tracer 'mmiotrace'
[    7.134151] Tracer 'mmiotrace' is not allowed on command line
[    7.134162] ------------[ cut here ]------------
[    7.134171] WARNING: CPU: 6 PID: 1 at kernel/trace/trace.c:5354 tracing_set_tracer+0x168/0x1b0
[    7.134171] Modules linked in:
[    7.134176] CPU: 6 PID: 1 Comm: swapper/0 Not tainted 4.13.0-1.mmiotracev3.fc26.x86_64 #1
[    7.134178] Hardware name: HP ProLiant DL380p Gen8, BIOS P70 09/18/2013
[    7.134179] task: ffff94b7310d8000 task.stack: ffffaadc83144000
[    7.134183] RIP: 0010:tracing_set_tracer+0x168/0x1b0
[    7.134185] RSP: 0000:ffffaadc83147da0 EFLAGS: 00010286
[    7.134187] RAX: 0000000000000031 RBX: ffffffffad04e340 RCX: 0000000000000000
[    7.134189] RDX: 0000000000000000 RSI: 0000000000000002 RDI: 0000000000000246
[    7.134190] RBP: ffffaadc83147dc8 R08: 0000000000000031 R09: ffffffffad314f00
[    7.134191] R10: ffffaadc83147d30 R11: 0000000000000000 R12: 0000000000000000
[    7.134193] R13: ffffffffacc935f3 R14: ffffffffacc935f3 R15: ffffffffacef46c0
[    7.134196] FS:  0000000000000000(0000) GS:ffff94b732e00000(0000) knlGS:0000000000000000
[    7.134197] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    7.134199] CR2: 0000000000000000 CR3: 00000005f5e09000 CR4: 00000000000406e0
[    7.134201] Call Trace:
[    7.134210]  ? set_debug_rodata+0x17/0x17
[    7.134213]  register_tracer+0x1a1/0x1c2
[    7.134216]  ? stack_trace_init+0xb5/0xb5
[    7.134218]  init_mmio_trace+0x10/0x12
[    7.134222]  do_one_initcall+0x50/0x190
[    7.134225]  kernel_init_freeable+0x1a8/0x245
[    7.134231]  ? rest_init+0xc0/0xc0
[    7.134234]  kernel_init+0xe/0x101
[    7.134238]  ret_from_fork+0x25/0x30
[    7.134240] Code: ff e8 6d d0 ff ff 85 c0 41 89 c4 79 8e eb b7 80 bb 9b 00 00 00 00 0f 84 26 ff ff ff 4c 89 ee 48 c7 c7 58 1e c9 ac e8 99 a0 f8 ff <0f> ff eb 97 48 8b 35 9d 81 d7 00 ba ff ff ff ff 4c 89 ff e8 10
[    7.134279] ---[ end trace 978ff599b6e516bc ]---
Quoted 6 lines

Also, I'm currently at OSS in LA and hopefully I don't lose this patch. If you don't see anything by next Monday from me, feel free to send me a reminder ping.

-- Steve

Signature · 6 lines
Ziqian SUN (Zamir)
9F Raycom office (NAY)
Red Hat Software (Beijing) Co.,Ltd
IRC: zsun (internal and freenode)
Tel: +86 10 65627458
GPG : 1D86 6D4A 49CE 4BBD 72CF FCF5 D856 6E11 F2A0 525E
#4 lore raw
Quoted 73 lines · Ziqian SUN (Zamir)
Quoted 37 lines · Steven Rostedt
Quoted 18 lines · On Mon, 11 Sep 2017 14:31:34 +0800 "Ziqian SUN (Zamir)" <[email protected]> wrote:
Quoted 17 lines · Ziqian SUN (Zamir)

From: "Ziqian SUN (Zamir)" <[email protected]>

The mmiotrace tracer cannot be enabled with ftrace=mmiotrace in kernel commandline. With this patch, noboot is added to the tracer struct, and when system boot with a tracer that has noboot=true, it will print out a warning message and continue booting.

Signature · 7 lines
v1 -> v2 : remove unessential comment
v2 -> v3 : Use tracer struct instead of a separate list to store noboot
---
   kernel/trace/trace.c           | 7 +++++++
   kernel/trace/trace.h           | 2 ++
   kernel/trace/trace_mmiotrace.c | 1 +
   3 files changed, 10 insertions(+)
Quoted 15 lines

kernel/trace/trace.c static int tracing_set_tracer(struct · line 5365

Quoted 13 lines
kernel/trace/trace.c +4 -0
index 5360b7a..48c474f 100644
@@ -5358,6 +5358,13 @@ static int tracing_set_tracer(struct 
 trace_array *tr, const char *buf)
        if (t == tr->current_trace)
            goto out;
+    /* Some tracers won't work on kernel command line */
+    if (system_state < SYSTEM_RUNNING && t->noboot) {
+        pr_warn("Tracer '%s' is not allowed on command line, 
 ignored\n",
+            t->name);

I feel the core trace printed by WARN is not so meaningful, so I use pr_warn instead.

Quoted 6 lines

I'm fine with pr_warn, but I'm curious to what you mean by "not so meaningful"? A WARN() will cause a dump stack, which usually shows up as a bug in systems and more likely to be seen. But if someone is adding this to the kernel command line and it's not working, they should be looking for the tracer name within the dmesg anyway.

Following is the log I tried with WARN before. With WARN we have both the warning message, and the stack trace. The stack trace goes into stacing_set_tracer, but not showing mmiotrace. I mean, this is useful for debugging, but maybe confusing for users who is willing to report bugs but not a kernel geek. With pr_warn we already indicate mmiotrace is ignored, and is neat.

Just my cents.

[    7.134149] Starting tracer 'mmiotrace'
[    7.134151] Tracer 'mmiotrace' is not allowed on command line
[    7.134162] ------------[ cut here ]------------
[    7.134171] WARNING: CPU: 6 PID: 1 at kernel/trace/trace.c:5354 tracing_set_tracer+0x168/0x1b0 [    7.134171] Modules linked in:
[    7.134176] CPU: 6 PID: 1 Comm: swapper/0 Not tainted 4.13.0-1.mmiotracev3.fc26.x86_64 #1 [    7.134178] Hardware name: HP ProLiant DL380p Gen8, BIOS P70 09/18/2013
[    7.134179] task: ffff94b7310d8000 task.stack: ffffaadc83144000
[    7.134183] RIP: 0010:tracing_set_tracer+0x168/0x1b0
[    7.134185] RSP: 0000:ffffaadc83147da0 EFLAGS: 00010286
[    7.134187] RAX: 0000000000000031 RBX: ffffffffad04e340 RCX: 0000000000000000 [    7.134189] RDX: 0000000000000000 RSI: 0000000000000002 RDI: 0000000000000246 [    7.134190] RBP: ffffaadc83147dc8 R08: 0000000000000031 R09: ffffffffad314f00 [    7.134191] R10: ffffaadc83147d30 R11: 0000000000000000 R12: 0000000000000000 [    7.134193] R13: ffffffffacc935f3 R14: ffffffffacc935f3 R15: ffffffffacef46c0 [    7.134196] FS:  0000000000000000(0000) GS:ffff94b732e00000(0000) knlGS:0000000000000000 [    7.134197] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    7.134199] CR2: 0000000000000000 CR3: 00000005f5e09000 CR4: 00000000000406e0 [    7.134201] Call Trace:
[    7.134210]  ? set_debug_rodata+0x17/0x17
[    7.134213]  register_tracer+0x1a1/0x1c2
[    7.134216]  ? stack_trace_init+0xb5/0xb5

Quoted 12 lines

[    7.134218]  init_mmio_trace+0x10/0x12
[    7.134222]  do_one_initcall+0x50/0x190
[    7.134225]  kernel_init_freeable+0x1a8/0x245
[    7.134231]  ? rest_init+0xc0/0xc0
[    7.134234]  kernel_init+0xe/0x101
[    7.134238]  ret_from_fork+0x25/0x30
[    7.134240] Code: ff e8 6d d0 ff ff 85 c0 41 89 c4 79 8e eb b7 80 bb 9b 00 00 00 00 0f 84 26 ff ff ff 4c 89 ee 48 c7 c7 58 1e c9 ac e8 99 a0 f8 ff <0f> ff eb 97 48 8b 35 9d 81 d7 00 ba ff ff ff ff 4c 89 ff e8 10 [    7.134279] ---[ end trace 978ff599b6e516bc ]---

Quoted 3 lines

Also, I'm currently at OSS in LA and hopefully I don't lose this patch. If you don't see anything by next Monday from me, feel free to send me a reminder ping.

Ping for updates.

Signature · 6 lines
Ziqian SUN (Zamir)
9F Raycom office (NAY)
Red Hat Software (Beijing) Co.,Ltd
IRC: zsun (internal and freenode)
Tel: +86 10 65627458
GPG : 1D86 6D4A 49CE 4BBD 72CF FCF5 D856 6E11 F2A0 525E
maintainer
#5 lore raw
In reply to Ziqian SUN (Zamir)
Quoted 10 lines · On Tue, 12 Sep 2017 20:18:35 +0800 "Ziqian SUN (Zamir)" <[email protected]> wrote:
Quoted 9 lines
Quoted 2 lines

I feel the core trace printed by WARN is not so meaningful, so I use pr_warn instead.

I'm fine with pr_warn, but I'm curious to what you mean by "not so meaningful"? A WARN() will cause a dump stack, which usually shows up as a bug in systems and more likely to be seen. But if someone is adding this to the kernel command line and it's not working, they should be looking for the tracer name within the dmesg anyway.

Quoted 7 lines

Following is the log I tried with WARN before. With WARN we have both the warning message, and the stack trace. The stack trace goes into stacing_set_tracer, but not showing mmiotrace. I mean, this is useful for debugging, but maybe confusing for users who is willing to report bugs but not a kernel geek. With pr_warn we already indicate mmiotrace is ignored, and is neat.

I'm now leaning towards WARN() over pr_warn(). Sure it's more useful for a kernel geek and not an everyday user, but I will counter that an everyday user should most definitely not be enabling mmiotrace! Or other traces for that matter.

But then again, it's not really a bug (it's a feature), so I guess I'll just take this patch as is.

-- Steve