2015-12-29 13:38:11

by Matias Bjørling

[permalink] [raw]
Subject: [PATCH] Fix for 4.4-rc7

Hi Jens,

Sorry for sending a patch so late. I found a bug upon generic media
manager initialization, followed by bad block list traversal, leading
to a faulty out-of-bound memory access. Unfortunately, the media manager
is initialized just after device initialization. In most cases the
kernel will crash upon boot when a device that supports multiple
channels is available.

Could you please pick this one up?

Thanks,
Matias

Matias Bjørling (1):
lightnvm: wrong offset in bad blk lun calculation

drivers/lightnvm/gennvm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

--
2.1.4


2015-12-29 13:38:13

by Matias Bjørling

[permalink] [raw]
Subject: [PATCH] lightnvm: wrong offset in bad blk lun calculation

dev->nr_luns reports the total number of luns available in a device
while dev->luns_per_chnl is the number of luns per channel.

When multiple channels are available, the offset is calculated from a
channel and lun id into a linear array. As it multiplies with
the total number of luns, we go out of bound when channel id > 0 and
causes the kernel to panic when we read a protected kernel memory area.

Signed-off-by: Matias Bjørling <[email protected]>
---
drivers/lightnvm/gennvm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/lightnvm/gennvm.c b/drivers/lightnvm/gennvm.c
index 04aead4..12ddcaa 100644
--- a/drivers/lightnvm/gennvm.c
+++ b/drivers/lightnvm/gennvm.c
@@ -75,7 +75,7 @@ static int gennvm_block_bb(struct ppa_addr ppa, int nr_blocks, u8 *blks,
struct nvm_block *blk;
int i;

- lun = &gn->luns[(dev->nr_luns * ppa.g.ch) + ppa.g.lun];
+ lun = &gn->luns[(dev->luns_per_chnl * ppa.g.ch) + ppa.g.lun];

for (i = 0; i < nr_blocks; i++) {
if (blks[i] == 0)
--
2.1.4

2015-12-29 15:29:16

by Jens Axboe

[permalink] [raw]
Subject: Re: [PATCH] lightnvm: wrong offset in bad blk lun calculation

On 12/29/2015 06:37 AM, Matias Bjørling wrote:
> dev->nr_luns reports the total number of luns available in a device
> while dev->luns_per_chnl is the number of luns per channel.
>
> When multiple channels are available, the offset is calculated from a
> channel and lun id into a linear array. As it multiplies with
> the total number of luns, we go out of bound when channel id > 0 and
> causes the kernel to panic when we read a protected kernel memory area.

Added for 4.4, thanks.

--
Jens Axboe