drm/i915: Avoid potential sleep whilst holding spinlock
authorChris Wilson <chris@chris-wilson.co.uk>
Tue, 20 Oct 2009 16:29:16 +0000 (16:29 +0000)
committerEric Anholt <eric@anholt.net>
Thu, 5 Nov 2009 21:33:22 +0000 (13:33 -0800)
Miles Lane reported the following error:
2 locks held by cat/4179:
  #0:  (&p->lock){+.+.+.}, at: [<c10a3884>] seq_read+0x25/0x315
  #1:  (&dev_priv->mm.active_list_lock){+.+...}, at: [<c119a854>]
i915_batchbuffer_info+0x2b/0x124
Pid: 4179, comm: cat Not tainted 2.6.32-rc5-git1 #2
Call Trace:
  [<c104874f>] ? __debug_show_held_locks+0x1e/0x20
  [<c1023fb0>] __might_sleep+0xf0/0xf7
  [<c101c393>] kmap+0x17/0x58
  [<c119a8d6>] i915_batchbuffer_info+0xad/0x124
  [<c10a39bf>] seq_read+0x160/0x315
  [<c108fb8c>] ? rw_verify_area+0x98/0xbb
  [<c10a385f>] ? seq_read+0x0/0x315
  [<c1090331>] vfs_read+0x75/0xa9
  [<c10903f9>] sys_read+0x3b/0x5d
  [<c1002a8f>] sysenter_do_call+0x12/0x36

The fix is relatively simple, use the atomic variants of kmap() that
avoid the potential sleep.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Miles Lane <miles.lane@gmail.com>
Signed-off-by: Eric Anholt <eric@anholt.net>
drivers/gpu/drm/i915/i915_debugfs.c

index f8ce9a3a420de39c39c2f517f0c067cfda7669a5..26bf0552b3cb572cf3c03b7c63e3a338573eddad 100644 (file)
@@ -267,10 +267,10 @@ static void i915_dump_pages(struct seq_file *m, struct page **pages, int page_co
        uint32_t *mem;
 
        for (page = 0; page < page_count; page++) {
-               mem = kmap(pages[page]);
+               mem = kmap_atomic(pages[page], KM_USER0);
                for (i = 0; i < PAGE_SIZE; i += 4)
                        seq_printf(m, "%08x :  %08x\n", i, mem[i / 4]);
-               kunmap(pages[page]);
+               kunmap_atomic(pages[page], KM_USER0);
        }
 }