[v2 PATCH] btrfs: raid56: Use correct stolen pages to calculate P/Q

[Qu Wenruo <quwenruo@cn.fujitsu.com>]

In the following situation, scrub will calculate wrong parity to
overwrite correct one:

RAID5 full stripe:

Before
|     Dev 1      |     Dev  2     |     Dev 3     |
| Data stripe 1  | Data stripe 2  | Parity Stripe |
--------------------------------------------------- 0
| 0x0000 (Bad)   |     0xcdcd     |     0x0000    |
--------------------------------------------------- 4K
|     0xcdcd     |     0xcdcd     |     0x0000    |
...
|     0xcdcd     |     0xcdcd     |     0x0000    |
--------------------------------------------------- 64K

After scrubbing dev3 only:

|     Dev 1      |     Dev  2     |     Dev 3     |
| Data stripe 1  | Data stripe 2  | Parity Stripe |
--------------------------------------------------- 0
| 0xcdcd (Good)  |     0xcdcd     | 0xcdcd (Bad)  |
--------------------------------------------------- 4K
|     0xcdcd     |     0xcdcd     |     0x0000    |
...
|     0xcdcd     |     0xcdcd     |     0x0000    |
--------------------------------------------------- 64K

The calltrace of such corruption is as following:

scrub_bio_end_io_worker() get called for each extent read out
|- scriub_block_complete()
   |- Data extent csum mismatch
   |- scrub_handle_errored_block
      |- scrub_recheck_block()
         |- scrub_submit_raid56_bio_wait()
            |- raid56_parity_recover()

Now we have a rbio with correct data stripe 1 recovered.
Let's call it "good_rbio".

scrub_parity_check_and_repair()
|- raid56_parity_submit_scrub_rbio()
   |- lock_stripe_add()
   |  |- steal_rbio()
   |     |- Recovered data are steal from "good_rbio", stored into
   |        rbio->stripe_pages[]
   |        Now rbio->bio_pages[] are bad data read from disk.
   |- async_scrub_parity()
      |- scrub_parity_work() (delayed_call to scrub_parity_work)

scrub_parity_work()
|- raid56_parity_scrub_stripe()
   |- validate_rbio_for_parity_scrub()
      |- finish_parity_scrub()
         |- Recalculate parity using *BAD* pages in rbio->bio_pages[]
            So good parity is overwritten with *BAD* one

The fix is to introduce 2 new members, bad_ondisk_a/b, to struct
btrfs_raid_bio, to info scrub code to use correct data pages to
re-calculate parity.

Reported-by: Goffredo Baroncelli <kreijack@inwind.it>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
Changelog:
v2:
  Add a new array of page pointers to record mapped pages and unmap
  these pages at the end.

After fixing this already hard to trace bug, we are able to expose more
bugs now!

I see some page data corruption if run parallel scrub (full fs scrub) with
above disk layout, the possibility is about 20%.

Seems there is some race in steal_rbio(), not the scrub race I fixed in
previous patches.
No one is able to found it before, because scrub doesn't even work
correctly before.

At this pace, RAID56 will never be stable, what we can do is to fix one
bug and expose another bug hidden by previous bug.
  
---
 fs/btrfs/raid56.c | 42 ++++++++++++++++++++++++++++++++++++++----
 1 file changed, 38 insertions(+), 4 deletions(-)

diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c
index d016d4a..994dffe 100644
--- a/fs/btrfs/raid56.c
+++ b/fs/btrfs/raid56.c
@@ -133,6 +133,16 @@ struct btrfs_raid_bio {
 	/* second bad stripe (for raid6 use) */
 	int failb;
 
+	/*
+	 * For steal_rbio, we can steal recovered correct page,
+	 * but in finish_parity_scrub(), we still use bad on-disk
+	 * page to calculate parity.
+	 * Use these members to info finish_parity_scrub() to use
+	 * correct pages
+	 */
+	int bad_ondisk_a;
+	int bad_ondisk_b;
+
 	int scrubp;
 	/*
 	 * number of pages needed to represent the full
@@ -310,6 +320,12 @@ static void steal_rbio(struct btrfs_raid_bio *src, struct btrfs_raid_bio *dest)
 	if (!test_bit(RBIO_CACHE_READY_BIT, &src->flags))
 		return;
 
+	/* Record recovered stripe number */
+	if (src->faila != -1)
+		dest->bad_ondisk_a = src->faila;
+	if (src->failb != -1)
+		dest->bad_ondisk_b = src->failb;
+
 	for (i = 0; i < dest->nr_pages; i++) {
 		s = src->stripe_pages[i];
 		if (!s || !PageUptodate(s)) {
@@ -998,6 +1014,8 @@ static struct btrfs_raid_bio *alloc_rbio(struct btrfs_root *root,
 	rbio->stripe_npages = stripe_npages;
 	rbio->faila = -1;
 	rbio->failb = -1;
+	rbio->bad_ondisk_a = -1;
+	rbio->bad_ondisk_b = -1;
 	atomic_set(&rbio->refs, 1);
 	atomic_set(&rbio->error, 0);
 	atomic_set(&rbio->stripes_pending, 0);
@@ -2292,6 +2310,7 @@ static noinline void finish_parity_scrub(struct btrfs_raid_bio *rbio,
 {
 	struct btrfs_bio *bbio = rbio->bbio;
 	void *pointers[rbio->real_stripes];
+	struct page *mapped_pages[rbio->real_stripes];
 	DECLARE_BITMAP(pbitmap, rbio->stripe_npages);
 	int nr_data = rbio->nr_data;
 	int stripe;
@@ -2352,12 +2371,24 @@ static noinline void finish_parity_scrub(struct btrfs_raid_bio *rbio,
 		void *parity;
 		/* first collect one page from each data stripe */
 		for (stripe = 0; stripe < nr_data; stripe++) {
-			p = page_in_rbio(rbio, stripe, pagenr, 0);
+
+			/*
+			 * Use stolen recovered page other than bad
+			 * on disk pages
+			 */
+			if (stripe == rbio->bad_ondisk_a ||
+			    stripe == rbio->bad_ondisk_b)
+				p = rbio_stripe_page(rbio, stripe, pagenr);
+			else
+				p = page_in_rbio(rbio, stripe, pagenr, 0);
 			pointers[stripe] = kmap(p);
+			mapped_pages[stripe] = p;
 		}
 
 		/* then add the parity stripe */
-		pointers[stripe++] = kmap(p_page);
+		pointers[stripe] = kmap(p_page);
+		mapped_pages[stripe] = p_page;
+		stripe++;
 
 		if (q_stripe != -1) {
 
@@ -2365,7 +2396,9 @@ static noinline void finish_parity_scrub(struct btrfs_raid_bio *rbio,
 			 * raid6, add the qstripe and call the
 			 * library function to fill in our p/q
 			 */
-			pointers[stripe++] = kmap(q_page);
+			pointers[stripe] = kmap(q_page);
+			mapped_pages[stripe] = q_page;
+			stripe++;
 
 			raid6_call.gen_syndrome(rbio->real_stripes, PAGE_SIZE,
 						pointers);
@@ -2385,8 +2418,9 @@ static noinline void finish_parity_scrub(struct btrfs_raid_bio *rbio,
 			bitmap_clear(rbio->dbitmap, pagenr, 1);
 		kunmap(p);
 
+		/* Free mapped pages */
 		for (stripe = 0; stripe < rbio->real_stripes; stripe++)
-			kunmap(page_in_rbio(rbio, stripe, pagenr, 0));
+			kunmap(mapped_pages[stripe]);
 	}
 
 	__free_page(p_page);
-- 
2.7.4