* how to handle tracing .h loops
@ 2011-08-18 11:51 Ian Campbell
2011-08-18 11:57 ` Ian Campbell
0 siblings, 1 reply; 3+ messages in thread
From: Ian Campbell @ 2011-08-18 11:51 UTC (permalink / raw)
To: Steven Rostedt, Frederic Weisbecker, Ingo Molnar; +Cc: linux-kernel
[-- Attachment #1: Type: text/plain, Size: 2878 bytes --]
Hi,
I'm working on a patch which needs to add a #include <linux/highmem.h>
to <linux/skbuff.h>. This is causing a bunch of fallout in the trace
headers which I can't figure out the right solution to. Simplifying the
patch to just:
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 7b996ed..3f5ee74 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -29,6 +29,7 @@
#include <linux/rcupdate.h>
#include <linux/dmaengine.h>
#include <linux/hrtimer.h>
+#include <linux/highmem.h>
/* Don't change this without changing skb_csum_unnecessary! */
#define CHECKSUM_NONE 0
reproduces on an i386 defconfig (v3.1-rc2). (actually just including
interrupt.h is sufficient too).
The spew is:
In file included from include/trace/ftrace.h:296:0,
from include/trace/define_trace.h:96,
from include/trace/events/irq.h:150,
from include/linux/interrupt.h:23,
from arch/x86/include/asm/highmem.h:23,
from include/linux/highmem.h:33,
from include/linux/skbuff.h:32,
from kernel/audit.h:24,
from kernel/signal.c:38:
include/trace/events/irq.h: In function 'ftrace_raw_output_softirq':
include/trace/events/irq.h:87:1: error: 'HI_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: note: each undeclared identifier is reported only once for each function it appears in
include/trace/events/irq.h:87:1: error: 'TIMER_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'NET_TX_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'NET_RX_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'BLOCK_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'BLOCK_IOPOLL_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'TASKLET_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'SCHED_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'HRTIMER_SOFTIRQ' undeclared (first use in this function)
include/trace/events/irq.h:87:1: error: 'RCU_SOFTIRQ' undeclared (first use in this function)
There are more following that but I suspect they are just secondary
fallout so I've omitted them in the interests of brevity. Full log is
attached.
The reason I can't see a solution is that it seems to be that
trace/events/irq.h in included in linux/interrupt.h before the
definition of the enum with all those values in it.
I looked in tip.git#perf/core (and master) but I can't see anything
which relates to this issue.
Clearly other places are including interrupt.h and highmem.h without
issue so what am I doing wrong here?
Thanks,
Ian.
[-- Attachment #2: log --]
[-- Type: application/octet-stream, Size: 8784 bytes --]
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: how to handle tracing .h loops
2011-08-18 11:51 how to handle tracing .h loops Ian Campbell
@ 2011-08-18 11:57 ` Ian Campbell
2011-08-18 13:24 ` Steven Rostedt
0 siblings, 1 reply; 3+ messages in thread
From: Ian Campbell @ 2011-08-18 11:57 UTC (permalink / raw)
To: Steven Rostedt; +Cc: Frederic Weisbecker, Ingo Molnar, linux-kernel
On Thu, 2011-08-18 at 12:51 +0100, Ian Campbell wrote:
> Clearly other places are including interrupt.h and highmem.h without
> issue so what am I doing wrong here?
I guess I should have look at the .c file instead of concentrating on
the .h's because this works:
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 7b996ed..3f5ee74 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -29,6 +29,7 @@
#include <linux/rcupdate.h>
#include <linux/dmaengine.h>
#include <linux/hrtimer.h>
+#include <linux/highmem.h>
/* Don't change this without changing skb_csum_unnecessary! */
#define CHECKSUM_NONE 0
diff --git a/kernel/signal.c b/kernel/signal.c
index 291c970..ff432ec 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -30,6 +30,7 @@
#include <linux/nsproxy.h>
#define CREATE_TRACE_POINTS
#include <trace/events/signal.h>
+#undef CREATE_TRACE_POINTS
#include <asm/param.h>
#include <asm/uaccess.h>
But that seems odd (noone else does it). Perhaps I should instead move
that include to last in the file? i.e.:
diff --git a/kernel/signal.c b/kernel/signal.c
index 291c970..d3cd4e7 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -28,8 +28,6 @@
#include <linux/freezer.h>
#include <linux/pid_namespace.h>
#include <linux/nsproxy.h>
-#define CREATE_TRACE_POINTS
-#include <trace/events/signal.h>
#include <asm/param.h>
#include <asm/uaccess.h>
@@ -37,6 +35,9 @@
#include <asm/siginfo.h>
#include "audit.h" /* audit_signal_info() */
+#define CREATE_TRACE_POINTS
+#include <trace/events/signal.h>
+
/*
* SLAB caches for signal bits.
*/
Ian.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: how to handle tracing .h loops
2011-08-18 11:57 ` Ian Campbell
@ 2011-08-18 13:24 ` Steven Rostedt
0 siblings, 0 replies; 3+ messages in thread
From: Steven Rostedt @ 2011-08-18 13:24 UTC (permalink / raw)
To: Ian Campbell; +Cc: Frederic Weisbecker, Ingo Molnar, linux-kernel
On Thu, 2011-08-18 at 12:57 +0100, Ian Campbell wrote:
> On Thu, 2011-08-18 at 12:51 +0100, Ian Campbell wrote:
> > Clearly other places are including interrupt.h and highmem.h without
> > issue so what am I doing wrong here?
>
> I guess I should have look at the .c file instead of concentrating on
> the .h's because this works:
>
> diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
> index 7b996ed..3f5ee74 100644
> --- a/include/linux/skbuff.h
> +++ b/include/linux/skbuff.h
> @@ -29,6 +29,7 @@
> #include <linux/rcupdate.h>
> #include <linux/dmaengine.h>
> #include <linux/hrtimer.h>
> +#include <linux/highmem.h>
>
> /* Don't change this without changing skb_csum_unnecessary! */
> #define CHECKSUM_NONE 0
> diff --git a/kernel/signal.c b/kernel/signal.c
> index 291c970..ff432ec 100644
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -30,6 +30,7 @@
> #include <linux/nsproxy.h>
> #define CREATE_TRACE_POINTS
> #include <trace/events/signal.h>
> +#undef CREATE_TRACE_POINTS
I think the better solution is to move this down below the asms. But as
this does work, I'm don't have any big preference against it.
>
> #include <asm/param.h>
> #include <asm/uaccess.h>
>
>
> But that seems odd (noone else does it). Perhaps I should instead move
> that include to last in the file? i.e.:
>
> diff --git a/kernel/signal.c b/kernel/signal.c
> index 291c970..d3cd4e7 100644
> --- a/kernel/signal.c
> +++ b/kernel/signal.c
> @@ -28,8 +28,6 @@
> #include <linux/freezer.h>
> #include <linux/pid_namespace.h>
> #include <linux/nsproxy.h>
> -#define CREATE_TRACE_POINTS
> -#include <trace/events/signal.h>
>
> #include <asm/param.h>
> #include <asm/uaccess.h>
> @@ -37,6 +35,9 @@
> #include <asm/siginfo.h>
> #include "audit.h" /* audit_signal_info() */
>
> +#define CREATE_TRACE_POINTS
> +#include <trace/events/signal.h>
> +
Yes, the trace events headers are suppose to be the last headers
included.
Either method works. Either undefine CREATE_TRACE_POINTS or move the
headers down. Only the C file should be depending on the header anyway.
Thanks!
-- Steve
> /*
> * SLAB caches for signal bits.
> */
>
> Ian.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2011-08-18 13:25 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-08-18 11:51 how to handle tracing .h loops Ian Campbell
2011-08-18 11:57 ` Ian Campbell
2011-08-18 13:24 ` Steven Rostedt
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®