rslib: Fix decoding of shortened codes
authorFerdinand Blomqvist <ferdinand.blomqvist@gmail.com>
Thu, 20 Jun 2019 14:10:34 +0000 (17:10 +0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sun, 4 Aug 2019 07:33:22 +0000 (09:33 +0200)
[ Upstream commit 2034a42d1747fc1e1eeef2c6f1789c4d0762cb9c ]

The decoding of shortenend codes is broken. It only works as expected if
there are no erasures.

When decoding with erasures, Lambda (the error and erasure locator
polynomial) is initialized from the given erasure positions. The pad
parameter is not accounted for by the initialisation code, and hence
Lambda is initialized from incorrect erasure positions.

The fix is to adjust the erasure positions by the supplied pad.

Signed-off-by: Ferdinand Blomqvist <ferdinand.blomqvist@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lkml.kernel.org/r/20190620141039.9874-3-ferdinand.blomqvist@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
lib/reed_solomon/decode_rs.c

index 0ec3f25..8eed0f9 100644 (file)
@@ -99,9 +99,9 @@
        if (no_eras > 0) {
                /* Init lambda to be the erasure locator polynomial */
                lambda[1] = alpha_to[rs_modnn(rs,
-                                             prim * (nn - 1 - eras_pos[0]))];
+                                       prim * (nn - 1 - (eras_pos[0] + pad)))];
                for (i = 1; i < no_eras; i++) {
-                       u = rs_modnn(rs, prim * (nn - 1 - eras_pos[i]));
+                       u = rs_modnn(rs, prim * (nn - 1 - (eras_pos[i] + pad)));
                        for (j = i + 1; j > 0; j--) {
                                tmp = index_of[lambda[j - 1]];
                                if (tmp != nn) {