Re: Re: how to avoid lost trace records?

[Olaf Hering <olaf@aepfle.de>]

[ Sorry, this mutt process was suspended and I didnt notice ... ]

On Mon, Nov 22, George Dunlap wrote:

> Olaf,
> 
> Dang, 8 megs per cpu -- but I guess that's really not so much overhead
> on a big machine; and it's definitely worth getting around the lost
> records issue.  Send the T_INFO_PAGES patch to the list, and see what
> Keir thinks.

The change is as simple as this, for xen-unstable. 4.0 needs to add the
get_order_from_pages() for alloc_xenheap_pages() calls.

A dynamic trace buffer allocation, and also allowing dynamic mask as
cmdline option, should be added. I have added this line to trace.c to
allow an event mask during booting:

integer_param("tbuf_event_mask", tb_event_mask);

Olaf

---
 xen/common/trace.c |    3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

--- xen-unstable.hg-4.1.22415.orig/xen/common/trace.c
+++ xen-unstable.hg-4.1.22415/xen/common/trace.c
@@ -48,7 +48,8 @@ integer_param("tbuf_size", opt_tbuf_size
 
 /* Pointers to the meta-data objects for all system trace buffers */
 static struct t_info *t_info;
-#define T_INFO_PAGES 2  /* Size fixed at 2 pages for now. */
+#define T_INFO_PAGES_ORDER 1  /* Size fixed at 2 pages for now. */
+#define T_INFO_PAGES (1 << T_INFO_PAGES_ORDER)
 #define T_INFO_SIZE ((T_INFO_PAGES)*(PAGE_SIZE))
 static DEFINE_PER_CPU_READ_MOSTLY(struct t_buf *, t_bufs);
 static DEFINE_PER_CPU_READ_MOSTLY(unsigned char *, t_data);