2002-08-20 17:39:07

by Marcus Alanen

Subject: [patch, 2.5] vmalloc.c error path fixes

I think there are some problems in vmalloc.c. The two first parts
of the diff fix a spinlock being held if an error occurs in
map_vm_area, and the last part fixes the error path of __vmalloc.

Perhaps somebody who knows more of the mm could verify this.

Marcus

===== mm/vmalloc.c 1.18 vs edited =====
--- 1.18/mm/vmalloc.c Mon Aug 5 22:05:22 2002
+++ edited/mm/vmalloc.c Mon Aug 19 00:37:40 2002
@@ -153,15 +153,20 @@
unsigned long address = VMALLOC_VMADDR(area->addr);
unsigned long end = address + (area->size-PAGE_SIZE);
pgd_t *dir;
+ int err = 0;

dir = pgd_offset_k(address);
spin_lock(&init_mm.page_table_lock);
do {
pmd_t *pmd = pmd_alloc(&init_mm, dir, address);
- if (!pmd)
- return -ENOMEM;
- if (map_area_pmd(pmd, address, end - address, prot, pages))
- return -ENOMEM;
+ if (!pmd) {
+ err = -ENOMEM;
+ break;
+ }
+ if (map_area_pmd(pmd, address, end - address, prot, pages)) {
+ err = -ENOMEM;
+ break;
+ }

address = (address + PGDIR_SIZE) & PGDIR_MASK;
dir++;
@@ -169,7 +174,7 @@

spin_unlock(&init_mm.page_table_lock);
flush_cache_all();
- return 0;
+ return err;
}


@@ -379,14 +384,20 @@

area->nr_pages = nr_pages;
area->pages = pages = kmalloc(array_size, (gfp_mask & ~__GFP_HIGHMEM));
- if (!area->pages)
+ if (!area->pages) {
+ remove_vm_area(area->addr);
+ kfree(area);
return NULL;
+ }
memset(area->pages, 0, array_size);

for (i = 0; i < area->nr_pages; i++) {
area->pages[i] = alloc_page(gfp_mask);
- if (unlikely(!area->pages[i]))
+ if (unlikely(!area->pages[i])) {
+ /* Successfully allocated i pages, free them in __vunmap() */
+ area->nr_pages = i;
goto fail;
+ }
}

if (map_vm_area(area, prot, &pages))

--
Marcus Alanen * Embedded Systems Laboratory * http://www.eslab.cs.abo.fi/
[email protected]

2002-08-20 18:33:31

by Andrew Morton

Subject: Re: [patch, 2.5] vmalloc.c error path fixes

Marcus Alanen wrote:
>
> I think there are some problems in vmalloc.c. The two first parts
> of the diff fix a spinlock being held if an error occurs in
> map_vm_area, and the last part fixes the error path of __vmalloc.

Certainly agree on the first chunk. Listen to the old guy: "Thou
shalt not have more than one return statement per function".

Second chunk looks good too, but perhaps there's another way of doing
it. The 2.5.31 code just tossed it all at vfree. Perhaps hch can
comment?

2002-08-23 19:07:25

by Christoph Hellwig

Subject: Re: [patch, 2.5] vmalloc.c error path fixes

On Tue, Aug 20, 2002 at 11:35:33AM -0700, Andrew Morton wrote:
> > of the diff fix a spinlock being held if an error occurs in
> > map_vm_area, and the last part fixes the error path of __vmalloc.
>
> Certainly agree on the first chunk. Listen to the old guy: "Thou
> shalt not have more than one return statement per function".
>
> Second chunk looks good too, but perhaps there's another way of doing
> it. The 2.5.31 code just tossed it all at vfree. Perhaps hch can
> comment?

The code looks fine - I missed that error checking. A little nitpick:
either remove the comment inn the last chunk as it's obvious what we
are doing or at least make it fit an ansi terminal.