[Lse-tech] Re: Bug in last night's bk test

[Andrew M. <ak...@di...>]

Badari Pulavarty wrote:
> 
> >
> > ...
> > Until that's done you'll need to set BIO_MAX_PAGES to 16 in
> > include/linux/bio.h
> >
> 
> I am little confused here. I thought IPS driver can handle 64K IO.
> Infact, IPS_MAX_SG is set to 17. So it should be able to handle 68K.
> I have been told that it can handle more than that.. but for some
> reason it was set to 17.
> 
> Paul, what kernel are u running ? 2.5.38 ?
> 

Current bitkeeper has 

#define BIO_MAX_PAGES           (256)

That's a megabyte.  It works fine with mpage.c.  But direct-io.c
is still using BIO_MAX_PAGES.  It really is building 1 megabyte
BIOs, which will break just about every device out there.

I think we just ask Linus to do the below until we get it fixed up?


--- 2.5.38-bk2/fs/direct-io.c~direct-io-size	Mon Sep 23 16:12:25 2002
+++ 2.5.38-bk2-akpm/fs/direct-io.c	Mon Sep 23 16:12:47 2002
@@ -26,7 +26,7 @@
  * The largest-sized BIO which this code will assemble, in bytes.  Set this
  * to PAGE_SIZE if your drivers are broken.
  */
-#define DIO_BIO_MAX_SIZE BIO_MAX_SIZE
+#define DIO_BIO_MAX_SIZE (16*1024)
 
 /*
  * How many user pages to map in one call to get_user_pages().  This determines

.