lkml.org 
[lkml]   [1999]   [Apr]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [patch] fix for cc1 Out of memory [Re: [TESTCASE] `Out of memory for cc1']


I was looking at the shared memory code, and just basically I cannot
convince myself that it can be reliable. It has all these hacky things in
it to bypass the normal mm layer logic, and I would not be at all
surprised if it has serious problems when swapping. Whether this could
explain any problems or not is unclear, but it does need to be cleaned up:
at least enough that I can convince myself that it could possibly work..

Didn't somebody (Andrea?) have a good test-program for shared memory under
low memory?

I have a truly cheezy patch that basically just rips out completely the
mm-bypass code, but I've so far only verified that it compiles. It makes
me feel happier about the shm code, but I don't have anything at _all_
that uses shm (I've even lost my xquake install in one of my machine
upgrades). Anybody interested in checking the patch out and seeing what
needs to be fixed up?

I bet there are things I missed - even though most of what I did was just
to remove a lot of dubious code and rely on the generic mm layer doing the
right thing instead.. But having _no_ testing it just basically is not
very likely to work on the first try.

Andrea, Stephen, any comments/interest?

This should be further fixed to use the page cache etc etc (and remove the
"attach" list that shouldn't be needed any more and other cleanups).
Consider this a weak first step.

Linus

-----
--- v2.2.5/linux/ipc/shm.c Thu Dec 31 10:29:03 1998
+++ linux/ipc/shm.c Mon Apr 12 14:27:26 1999
@@ -24,7 +24,8 @@
static void killseg (int id);
static void shm_open (struct vm_area_struct *shmd);
static void shm_close (struct vm_area_struct *shmd);
-static pte_t shm_swap_in(struct vm_area_struct *, unsigned long, unsigned long);
+static unsigned long shm_nopage(struct vm_area_struct *, unsigned long, int);
+static int shm_swapout(struct vm_area_struct *, struct page *);

static int shm_tot = 0; /* total number of shared memory pages */
static int shm_rss = 0; /* number of shared memory pages that are in memory */
@@ -364,10 +365,10 @@
NULL, /* protect */
NULL, /* sync */
NULL, /* advise */
- NULL, /* nopage (done with swapin) */
+ shm_nopage, /* nopage */
NULL, /* wppage */
- NULL, /* swapout (hardcoded right now) */
- shm_swap_in /* swapin */
+ shm_swapout, /* swapout */
+ NULL /* swapin */
};

/* Insert shmd into the list shp->attaches */
@@ -393,11 +394,7 @@
*/
static int shm_map (struct vm_area_struct *shmd)
{
- pgd_t *page_dir;
- pmd_t *page_middle;
- pte_t *page_table;
- unsigned long tmp, shm_sgn;
- int error;
+ unsigned long tmp;

/* clear old mappings */
do_munmap(shmd->vm_start, shmd->vm_end - shmd->vm_start);
@@ -411,30 +408,7 @@
insert_vm_struct(current->mm, shmd);
merge_segments(current->mm, shmd->vm_start, shmd->vm_end);

- /* map page range */
- error = 0;
- shm_sgn = shmd->vm_pte +
- SWP_ENTRY(0, (shmd->vm_offset >> PAGE_SHIFT) << SHM_IDX_SHIFT);
- flush_cache_range(shmd->vm_mm, shmd->vm_start, shmd->vm_end);
- for (tmp = shmd->vm_start;
- tmp < shmd->vm_end;
- tmp += PAGE_SIZE, shm_sgn += SWP_ENTRY(0, 1 << SHM_IDX_SHIFT))
- {
- page_dir = pgd_offset(shmd->vm_mm,tmp);
- page_middle = pmd_alloc(page_dir,tmp);
- if (!page_middle) {
- error = -ENOMEM;
- break;
- }
- page_table = pte_alloc(page_middle,tmp);
- if (!page_table) {
- error = -ENOMEM;
- break;
- }
- set_pte(page_table, __pte(shm_sgn));
- }
- flush_tlb_range(shmd->vm_mm, shmd->vm_start, shmd->vm_end);
- return error;
+ return 0;
}

/*
@@ -615,44 +589,46 @@
}

/*
+ * Enter the shm page into the SHM data structures.
+ *
+ * The way "nopage" is done, we don't actually have to
+ * do anything here: nopage will have filled in the shm
+ * data structures already, and shm_swap_out() will just
+ * work off them..
+ */
+static int shm_swapout(struct vm_area_struct * vma, struct page * page)
+{
+ return 0;
+}
+
+/*
* page not present ... go through shm_pages
*/
-static pte_t shm_swap_in(struct vm_area_struct * shmd, unsigned long offset, unsigned long code)
+static unsigned long shm_nopage(struct vm_area_struct * shmd, unsigned long address, int no_share)
{
pte_t pte;
struct shmid_kernel *shp;
unsigned int id, idx;

- id = SWP_OFFSET(code) & SHM_ID_MASK;
+ id = SWP_OFFSET(shmd->vm_pte) & SHM_ID_MASK;
+ idx = (address - shmd->vm_start + shmd->vm_offset) >> PAGE_SHIFT;
+
#ifdef DEBUG_SHM
- if (id != (SWP_OFFSET(shmd->vm_pte) & SHM_ID_MASK)) {
- printk ("shm_swap_in: code id = %d and shmd id = %ld differ\n",
- id, SWP_OFFSET(shmd->vm_pte) & SHM_ID_MASK);
- return BAD_PAGE;
- }
if (id > max_shmid) {
- printk ("shm_swap_in: id=%d too big. proc mem corrupted\n", id);
- return BAD_PAGE;
+ printk ("shm_nopage: id=%d too big. proc mem corrupted\n", id);
+ return 0;
}
#endif
shp = shm_segs[id];

#ifdef DEBUG_SHM
if (shp == IPC_UNUSED || shp == IPC_NOID) {
- printk ("shm_swap_in: id=%d invalid. Race.\n", id);
- return BAD_PAGE;
- }
-#endif
- idx = (SWP_OFFSET(code) >> SHM_IDX_SHIFT) & SHM_IDX_MASK;
-#ifdef DEBUG_SHM
- if (idx != (offset >> PAGE_SHIFT)) {
- printk ("shm_swap_in: code idx = %u and shmd idx = %lu differ\n",
- idx, offset >> PAGE_SHIFT);
- return BAD_PAGE;
+ printk ("shm_nopage: id=%d invalid. Race.\n", id);
+ return 0;
}
if (idx >= shp->shm_npages) {
- printk ("shm_swap_in : too large page index. id=%d\n", id);
- return BAD_PAGE;
+ printk ("shm_nopage : too large page index. id=%d\n", id);
+ return 0;
}
#endif

@@ -661,7 +637,7 @@
unsigned long page = get_free_page(GFP_KERNEL);
if (!page) {
oom(current);
- return BAD_PAGE;
+ return 0;
}
pte = __pte(shp->shm_pages[idx]);
if (pte_present(pte)) {
@@ -687,7 +663,7 @@
done: /* pte_val(pte) == shp->shm_pages[idx] */
current->min_flt++;
atomic_inc(&mem_map[MAP_NR(pte_page(pte))].count);
- return pte_modify(pte, shmd->vm_page_prot);
+ return pte_page(pte);
}

/*
@@ -700,7 +676,6 @@
{
pte_t page;
struct shmid_kernel *shp;
- struct vm_area_struct *shmd;
unsigned long swap_nr;
unsigned long id, idx;
int loop = 0;
@@ -742,61 +717,6 @@
swap_free (swap_nr);
return 0;
}
- if (shp->attaches)
- for (shmd = shp->attaches; ; ) {
- do {
- pgd_t *page_dir;
- pmd_t *page_middle;
- pte_t *page_table, pte;
- unsigned long tmp;
-
- if ((SWP_OFFSET(shmd->vm_pte) & SHM_ID_MASK) != id) {
- printk ("shm_swap: id=%ld does not match shmd->vm_pte.id=%ld\n",
- id, SWP_OFFSET(shmd->vm_pte) & SHM_ID_MASK);
- continue;
- }
- tmp = shmd->vm_start + (idx << PAGE_SHIFT) - shmd->vm_offset;
- if (!(tmp >= shmd->vm_start && tmp < shmd->vm_end))
- continue;
- page_dir = pgd_offset(shmd->vm_mm,tmp);
- if (pgd_none(*page_dir) || pgd_bad(*page_dir)) {
- printk("shm_swap: bad pgtbl! id=%ld start=%lx idx=%ld\n",
- id, shmd->vm_start, idx);
- pgd_clear(page_dir);
- continue;
- }
- page_middle = pmd_offset(page_dir,tmp);
- if (pmd_none(*page_middle) || pmd_bad(*page_middle)) {
- printk("shm_swap: bad pgmid! id=%ld start=%lx idx=%ld\n",
- id, shmd->vm_start, idx);
- pmd_clear(page_middle);
- continue;
- }
- page_table = pte_offset(page_middle,tmp);
- pte = *page_table;
- if (!pte_present(pte))
- continue;
- if (pte_young(pte)) {
- set_pte(page_table, pte_mkold(pte));
- continue;
- }
- if (pte_page(pte) != pte_page(page))
- printk("shm_swap_out: page and pte mismatch %lx %lx\n",
- pte_page(pte),pte_page(page));
- flush_cache_page(shmd, tmp);
- set_pte(page_table,
- __pte(shmd->vm_pte + SWP_ENTRY(0, idx << SHM_IDX_SHIFT)));
- atomic_dec(&mem_map[MAP_NR(pte_page(pte))].count);
- if (shmd->vm_mm->rss > 0)
- shmd->vm_mm->rss--;
- flush_tlb_page(shmd, tmp);
- /* continue looping through the linked list */
- } while (0);
- shmd = shmd->vm_next_share;
- if (!shmd)
- break;
- }
-
if (atomic_read(&mem_map[MAP_NR(pte_page(page))].count) != 1)
goto check_table;
shp->shm_pages[idx] = swap_nr;


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

\
 
 \ /
  Last update: 2005-03-22 13:51    [from the cache]
©2003-2014 Jasper Spaans. Advertise on this site