[uml-devel] [PATCH] Fix for annoying problem with hostfs readdir problems.

[uml-devel] [PATCH] Fix for annoying problem with hostfs readdir problems.

[Oleg Drokin]

[-- Attachment #1: Type: text/plain, Size: 364 bytes --]

Hello!

   So the annoying problem with readdir not working on hostfs (on 2.6 host
   kernel only for me) was finally hunted to memory allocation
   problem during opendir.

   The patch below is a working and tested attempt at fixing it.
   It is againt 2.4.20-something, but is pretty straightforward and
   probably should apply everywhere else.

Bye,
    Oleg

[-- Attachment #2: malloc_fix.diff --]
[-- Type: text/plain, Size: 1510 bytes --]

--- uml-2.4/arch/um/kernel/process_kern.c.orig	2004-04-15 23:26:49.000000000 +0300
+++ uml-2.4/arch/um/kernel/process_kern.c	2004-04-15 23:49:38.312250576 +0300
@@ -16,6 +16,7 @@
 #include "linux/module.h"
 #include "linux/init.h"
 #include "linux/capability.h"
+#include "linux/vmalloc.h"
 #include "asm/unistd.h"
 #include "asm/mman.h"
 #include "asm/segment.h"
@@ -279,6 +280,11 @@ void *um_kmalloc(int size)
 	return(kmalloc(size, GFP_KERNEL));
 }
 
+void *um_vmalloc(int size)
+{
+	return(vmalloc(size));
+}
+
 void *um_kmalloc_atomic(int size)
 {
 	return(kmalloc(size, GFP_ATOMIC));
--- uml-2.4/arch/um/main.c.orig	2004-04-15 23:58:28.786606216 +0300
+++ uml-2.4/arch/um/main.c	2004-04-15 23:53:32.706617216 +0300
@@ -161,7 +161,7 @@ extern void *__real_malloc(int);
 void *__wrap_malloc(int size)
 {
 	if(CAN_KMALLOC())
-		return(um_kmalloc(size));
+		return(um_vmalloc(size));
 	else
 		return(__real_malloc(size));
 }
@@ -179,7 +179,7 @@ extern void __real_free(void *);
 
 void __wrap_free(void *ptr)
 {
-	if(CAN_KMALLOC()) kfree(ptr);
+	if(CAN_KMALLOC()) vfree(ptr);
 	else __real_free(ptr);
 }
 
--- uml-2.4/arch/um/include/user.h.orig	2004-04-15 23:41:57.000000000 +0300
+++ uml-2.4/arch/um/include/user.h	2004-04-15 23:41:34.000000000 +0300
@@ -12,6 +12,7 @@ extern void schedule(void);
 extern void *um_kmalloc(int size);
 extern void *um_kmalloc_atomic(int size);
 extern void kfree(void *ptr);
+extern void vfree(void *ptr);
 extern int in_aton(char *str);
 extern int open_gdb_chan(void);
 

[uml-devel] Re: [PATCH] Fix for annoying problem with hostfs readdir problems.

[Oleg Drokin]

[-- Attachment #1: Type: text/plain, Size: 608 bytes --]

Hello!

On Fri, Apr 16, 2004 at 12:03:18AM +0300, Oleg Drokin wrote:
>    So the annoying problem with readdir not working on hostfs (on 2.6 host
>    kernel only for me) was finally hunted to memory allocation
>    problem during opendir.
> 
>    The patch below is a working and tested attempt at fixing it.
>    It is againt 2.4.20-something, but is pretty straightforward and
>    probably should apply everywhere else.

This is corrected version that sets errno on malloc failure and as a bonus -
um_vmalloc is now also declared in header file and therefore no edtra warning in
main.c ;)

Bye,
    Oleg

[-- Attachment #2: malloc_fix.diff --]
[-- Type: text/plain, Size: 2043 bytes --]

--- uml-2.4/arch/um/kernel/process_kern.c.orig	2004-04-15 23:26:49.000000000 +0300
+++ uml-2.4/arch/um/kernel/process_kern.c	2004-04-15 23:49:38.312250576 +0300
@@ -16,6 +16,7 @@
 #include "linux/module.h"
 #include "linux/init.h"
 #include "linux/capability.h"
+#include "linux/vmalloc.h"
 #include "asm/unistd.h"
 #include "asm/mman.h"
 #include "asm/segment.h"
@@ -279,6 +280,11 @@ void *um_kmalloc(int size)
 	return(kmalloc(size, GFP_KERNEL));
 }
 
+void *um_vmalloc(int size)
+{
+	return(vmalloc(size));
+}
+
 void *um_kmalloc_atomic(int size)
 {
 	return(kmalloc(size, GFP_ATOMIC));
--- uml-2.4/arch/um/main.c.orig	2004-04-15 23:58:28.786606216 +0300
+++ uml-2.4/arch/um/main.c	2004-04-16 00:35:35.074159064 +0300
@@ -8,6 +8,7 @@
 #include <stdlib.h>
 #include <string.h>
 #include <signal.h>
+#include <errno.h>
 #include <sys/resource.h>
 #include <sys/mman.h>
 #include <sys/user.h>
@@ -160,10 +161,16 @@ extern void *__real_malloc(int);
 
 void *__wrap_malloc(int size)
 {
-	if(CAN_KMALLOC())
-		return(um_kmalloc(size));
-	else
+	if(CAN_KMALLOC()) {
+		/* glibc people insist that if malloc fails, errno should be
+		   set by malloc as well. So we do. */
+		void *ret = um_vmalloc(size);
+		if (!ret)
+			errno = ENOMEM;
+		return(ret);
+	} else {
 		return(__real_malloc(size));
+	}
 }
 
 void *__wrap_calloc(int n, int size)
@@ -179,7 +186,7 @@ extern void __real_free(void *);
 
 void __wrap_free(void *ptr)
 {
-	if(CAN_KMALLOC()) kfree(ptr);
+	if(CAN_KMALLOC()) vfree(ptr);
 	else __real_free(ptr);
 }
 
--- uml-2.4/arch/um/include/user.h.orig	2004-04-15 23:41:57.000000000 +0300
+++ uml-2.4/arch/um/include/user.h	2004-04-16 00:36:24.806598584 +0300
@@ -10,8 +10,10 @@ extern void panic(const char *fmt, ...);
 extern int printk(const char *fmt, ...);
 extern void schedule(void);
 extern void *um_kmalloc(int size);
+extern void *um_vmalloc(int size);
 extern void *um_kmalloc_atomic(int size);
 extern void kfree(void *ptr);
+extern void vfree(void *ptr);
 extern int in_aton(char *str);
 extern int open_gdb_chan(void);
 

Re: [uml-devel] Re: [PATCH] Fix for annoying problem with hostfs readdir problems.

[BlaisorBlade]

Alle 23:46, giovedì 15 aprile 2004, Oleg Drokin ha scritto:
> Hello!
>
> On Fri, Apr 16, 2004 at 12:03:18AM +0300, Oleg Drokin wrote:
> >    So the annoying problem with readdir not working on hostfs (on 2.6
> > host kernel only for me) was finally hunted to memory allocation
> >    problem during opendir.

I've experienced it too onto 2.6 and someone else had the same problem - but 
that patch is not perfect. Using always vmalloc() instead of kmalloc() is 
wrong - vmalloc is slower, and Linus spoke against vmalloc(); it would also 
slow down everything using malloc().

Probably you should use vmalloc() only if size is > the size limit (if this is 
the problem, as I guess). When you come at free() it is a bit hard, but IIRC 
vmalloc() can return pointers inside a certain, known memory area, right? So 
it is possible to understand what function to call.

Bye
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id\x1470&alloc_id638&opÌk
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] Re: [PATCH] Fix for annoying problem with hostfs readdir problems.

[Oleg Drokin]

Hello!

On Sun, Apr 18, 2004 at 05:26:23PM +0200, BlaisorBlade wrote:
> > >    So the annoying problem with readdir not working on hostfs (on 2.6
> > > host kernel only for me) was finally hunted to memory allocation
> > >    problem during opendir.
> I've experienced it too onto 2.6 and someone else had the same problem - but 
> that patch is not perfect. Using always vmalloc() instead of kmalloc() is 
> wrong - vmalloc is slower, and Linus spoke against vmalloc(); it would also 
> slow down everything using malloc().

Well, speed is not of much an issue there, that function is only used for
glibc allocations. And that read_dir() is probably the only place that
causes glibc to allocate something.

> Probably you should use vmalloc() only if size is > the size limit (if this is 
> the problem, as I guess). When you come at free() it is a bit hard, but IIRC 
> vmalloc() can return pointers inside a certain, known memory area, right? So 
> it is possible to understand what function to call.

I am sure that all of the complexity you propose to implement won't pay off in
the end.

Bye,
    Oleg


-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel