From: Eric Moore Date: Sat, 29 Sep 2007 16:17:49 +0000 (-0600) Subject: [SCSI] mptlan: bug fix, only half the message frame is dma'd resulting in corruption X-Git-Tag: v2.6.24-rc1~1290^2~55 X-Git-Url: https://err.no/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=e1fc2b5167f9c3614099e55d3f49763a26a5a339;p=linux-2.6 [SCSI] mptlan: bug fix, only half the message frame is dma'd resulting in corruption NB = number of blocks. This represents the number of blocks to transfer. The block size is based on the message frame size provided in the ioc_facts. A value of zero indicates the entire message frame should be copied. This is two bit value. So by setting this to non-zero vaule, you increase performance by reducing amount of data needing to be dma'd. The value that is stored in ioc->ReqeustNB is sometimes a non-zero vaule, which creates a bug in mptlan, where not the entire message frame is getting transfer to firware, resulting in corruption. This fix sets the default to zero, thus entire message frame is copied. Signed-off-by: Eric Moore Signed-off-by: James Bottomley --- diff --git a/drivers/message/fusion/mptlan.c b/drivers/message/fusion/mptlan.c index b550471e30..8558322073 100644 --- a/drivers/message/fusion/mptlan.c +++ b/drivers/message/fusion/mptlan.c @@ -1224,6 +1224,8 @@ mpt_lan_post_receive_buckets(struct mpt_lan_priv *priv) } pRecvReq = (LANReceivePostRequest_t *) mf; + i = le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx); + mpt_dev->RequestNB[i] = 0; count = buckets; if (count > max) count = max;