> -----Original Message----- > From: santosh prasad nayak [mailto:santoshprasadnayak@xxxxxxxxx] > Sent: Saturday, March 10, 2012 12:20 AM > To: Rajesh Borundia > Cc: Sony Chacko; netdev; linux-kernel; kernel-janitors@xxxxxxxxxxxxxxx > Subject: Re: [PATCH 3/3] netxen: qlogic ethernet : Fix Endian Bug. > > On Fri, Mar 9, 2012 at 10:04 PM, Rajesh Borundia > <rajesh.borundia@xxxxxxxxxx> wrote: > > > >> -----Original Message----- > >> From: santosh nayak [mailto:santoshprasadnayak@xxxxxxxxx] > >> Sent: Saturday, March 03, 2012 9:18 PM > >> To: Sony Chacko > >> Cc: Rajesh Borundia; netdev; linux-kernel; kernel- > >> janitors@xxxxxxxxxxxxxxx; Santosh Nayak > >> Subject: [PATCH 3/3] netxen: qlogic ethernet : Fix Endian Bug. > >> > >> From: Santosh Nayak <santoshprasadnayak@xxxxxxxxx> > >> > >> Fix endian bug. > >> > >> Signed-off-by: Santosh Nayak <santoshprasadnayak@xxxxxxxxx> > >> --- > >> drivers/net/ethernet/qlogic/netxen/netxen_nic.h | 4 ++-- > >> drivers/net/ethernet/qlogic/netxen/netxen_nic_hw.c | 12 +++++++-- > --- > >> .../net/ethernet/qlogic/netxen/netxen_nic_main.c | 2 +- > >> 3 files changed, 10 insertions(+), 8 deletions(-) > >> > >> diff --git a/drivers/net/ethernet/qlogic/netxen/netxen_nic.h > >> b/drivers/net/ethernet/qlogic/netxen/netxen_nic.h > >> index 2eeac32..b5de8a7 100644 > >> --- a/drivers/net/ethernet/qlogic/netxen/netxen_nic.h > >> +++ b/drivers/net/ethernet/qlogic/netxen/netxen_nic.h > >> @@ -954,7 +954,7 @@ typedef struct nx_mac_list_s { > >> > >> struct nx_vlan_ip_list { > >> struct list_head list; > >> - u32 ip_addr; > >> + __be32 ip_addr; > >> }; > >> > >> /* > >> @@ -1780,7 +1780,7 @@ int netxen_process_rcv_ring(struct > >> nx_host_sds_ring *sds_ring, int max); > >> void netxen_p3_free_mac_list(struct netxen_adapter *adapter); > >> int netxen_config_intr_coalesce(struct netxen_adapter *adapter); > >> int netxen_config_rss(struct netxen_adapter *adapter, int enable); > >> -int netxen_config_ipaddr(struct netxen_adapter *adapter, u32 ip, > int > >> cmd); > >> +int netxen_config_ipaddr(struct netxen_adapter *adapter, __be32 ip, > >> int cmd); > >> int netxen_linkevent_request(struct netxen_adapter *adapter, int > >> enable); > >> void netxen_advert_link_change(struct netxen_adapter *adapter, int > >> linkup); > >> void netxen_pci_camqm_read_2M(struct netxen_adapter *, u64, u64 *); > >> diff --git a/drivers/net/ethernet/qlogic/netxen/netxen_nic_hw.c > >> b/drivers/net/ethernet/qlogic/netxen/netxen_nic_hw.c > >> index 6f37470..0f81287 100644 > >> --- a/drivers/net/ethernet/qlogic/netxen/netxen_nic_hw.c > >> +++ b/drivers/net/ethernet/qlogic/netxen/netxen_nic_hw.c > >> @@ -909,10 +909,11 @@ int netxen_config_rss(struct netxen_adapter > >> *adapter, int enable) > >> return rv; > >> } > >> > >> -int netxen_config_ipaddr(struct netxen_adapter *adapter, u32 ip, > int > >> cmd) > >> +int netxen_config_ipaddr(struct netxen_adapter *adapter, __be32 ip, > >> int cmd) > >> { > >> nx_nic_req_t req; > >> u64 word; > >> + u64 ip_addr; > >> int rv; > >> > >> memset(&req, 0, sizeof(nx_nic_req_t)); > >> @@ -922,7 +923,8 @@ int netxen_config_ipaddr(struct netxen_adapter > >> *adapter, u32 ip, int cmd) > >> req.req_hdr = cpu_to_le64(word); > >> > >> req.words[0] = cpu_to_le64(cmd); > >> - req.words[1] = cpu_to_le64(ip); > >> + ip_addr = be32_to_cpu(ip); > >> + *(__be64 *)&req.words[1] = cpu_to_be64(ip_addr); > > > > > > Adapter requires ip value in big endian stored at lower 32 bit > address. > > The cpu_to_be64 will swap the lower 32 bit with higher 32 bit and > adapter > > Will get incorrect ip addr. Instead u can do. > > > > U32 *ip_addr; > > ip_addr = (u32 *)&req.words[1]; > > *ip_addr = ip; > > > 1. It looks incomplete. > In the function call "netxen_send_cmd_descs" we have to pass "&req" > as > 2nd argument not "ipaddr". I should have sent a patch. This piece of code was just to show how to copy ip addr in req.words[1]. > > 2. Your above suggestion is with assumption that the data type of 2nd > argument "ip" > in "netxen_config_ipaddr()" is still "u32". This is not true. > > Some days back you suggested to change the data type to "__be32". > In the present patch > the "ip" is in "__be32" format i.e already in Big endian format > as per requirement. > We need to only convert this 32 bit to 64 bit. There are two > ways: > No I did not assume that ip is u32, ip is still __be32(ip value is in form of big endian) though I should have mentioned it explicitly. But the ip value should be copied to lower 32 bit of req.words[1]. > a. *(__be64 *)&req.words[1] = ip; // auto conversion In big endian machine MSB is copied into MSB first. So ip will get copied into higher 32 bit of req.words[1] but adapter requires it in lower 32 bit. > > b. *(__be64 *)&req.words[1] = cpu_to_be64(be32_to_cpu(ip)); > // explicit conversion. > If you follow second cpu_to_be64 it will swap lower 32 bit of ip with higher 32 bit in little endian machine. But adapter requires it in lower 32 bit. Simple solution to copy ip in req.words[1] could be memcpy(&req.words[1], &ip, sizeof(u32)); > > Please correct me if I am wrong. > > > regards > Santosh > > > > > > > > > >> > >> rv = netxen_send_cmd_descs(adapter, (struct cmd_desc_type0 > >> *)&req, 1); > >> if (rv != 0) { > >> @@ -1050,7 +1052,7 @@ int netxen_get_flash_mac_addr(struct > >> netxen_adapter *adapter, u64 *mac) > >> if (netxen_get_flash_block(adapter, offset, sizeof(u64), pmac) > == > >> -1) > >> return -1; > >> > >> - if (*mac == cpu_to_le64(~0ULL)) { > >> + if (*mac == ~0ULL) { > > > > *mac is in little endian format so compare it with cpu_to_le64. > > > >> > >> offset = NX_OLD_MAC_ADDR_OFFSET + > >> (adapter->portnum * sizeof(u64)); > >> @@ -1059,7 +1061,7 @@ int netxen_get_flash_mac_addr(struct > >> netxen_adapter *adapter, u64 *mac) > >> offset, sizeof(u64), pmac) == > -1) > >> return -1; > >> > >> - if (*mac == cpu_to_le64(~0ULL)) > >> + if (*mac == ~0ULL) > > > > *mac here is in little endian format so compare it with cpu_to_le64. > > > >> return -1; > >> } > >> return 0; > >> @@ -2178,7 +2180,7 @@ lock_try: > >> NX_WR_DUMP_REG(FLASH_ROM_WINDOW, adapter- > >ahw.pci_base0, > >> waddr); > >> raddr = FLASH_ROM_DATA + (fl_addr & 0x0000FFFF); > >> NX_RD_DUMP_REG(raddr, adapter->ahw.pci_base0, &val); > >> - *data_buff++ = cpu_to_le32(val); > >> + *data_buff++ = val; > > > > It should be cpu_to_le32 as it is returned to tool which requires > > output in little endian. > > > >> fl_addr += sizeof(val); > >> } > >> readl((void __iomem *)(adapter->ahw.pci_base0 + > >> NX_FLASH_SEM2_ULK)); > >> diff --git a/drivers/net/ethernet/qlogic/netxen/netxen_nic_main.c > >> b/drivers/net/ethernet/qlogic/netxen/netxen_nic_main.c > >> index 8dc4a134..70783b4 100644 > >> --- a/drivers/net/ethernet/qlogic/netxen/netxen_nic_main.c > >> +++ b/drivers/net/ethernet/qlogic/netxen/netxen_nic_main.c > >> @@ -818,7 +818,7 @@ netxen_check_options(struct netxen_adapter > >> *adapter) > >> adapter->driver_mismatch = 1; > >> return; > >> } > >> - ptr32[i] = cpu_to_le32(val); > >> + ptr32[i] = val; > > > > Here val should be in little endian (cpu_to_le32) as it will be > referenced by byte array to print serial number. > > > >> offset += sizeof(u32); > >> } > >> > >> -- > >> 1.7.4.4 > >> > > > > > > Sorry for Late reply. > > > > Rajesh > > -- To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
