Hi Arend,
I am getting a Data bus error when brcmsmac tries to read tsf_random in
brcms_c_wme_setparams. This does not happen when I add a printk
somewhere near that read of tsf_random or when I replace the read with a
constant. In b43 this code works, without this error for me, see
b43_qos_params_upload(). With the BCM4716 I do not have this problem. It
could be that this is caused by the broken PCIe controller in the BCM4716.
I am using brcmsmac from compat-wireless 2012-09-07, but I have also
tried the most recent brcmsmac code from wireless-testing. brcmsmac uses
the normal firmware from linux-firmware and I have also tried the new
firmware posted some days ago with the same result.
I am using a BCM43224 connected to the PCIe controller of the BCM4716.
Are there any special requirements for reading tsf_random under mips?
To make my BCM43224 work I use this patch [0], the "printk("dummy\n");"
is a hack to workaround the problem I am describing here.
Hauke
[0]:
https://dev.openwrt.org/browser/trunk/package/mac80211/patches/850-brcmsmac-add-support-for-BCM43224.patch
root@OpenWrt:/# iw phy phy0 interface add wlan0 type managed
root@OpenWrt:/# ifconfig wlan0 up
[ 34.828000] Data bus error, epc == 8014a890, ra == 801ba3c4
[ 34.828000] Oops[#1]:
[ 34.828000] Cpu 0
[ 34.828000] $ 0 : 00000000 1100cc00 00000001 83994800
[ 34.828000] $ 4 : a800065a 00000820 00000000 00000002
[ 34.828000] $ 8 : 00000020 801392f0 24189612 6c604830
[ 34.828000] $12 : 00000000 00000001 00000000 ffffff00
[ 34.828000] $16 : 83992e80 0000065a 00000003 00000020
[ 34.828000] $20 : 00000000 83992e80 82e3a0e4 82e5b458
[ 34.828000] $24 : 00000000 801ba2e8
[ 34.828000] $28 : 82e98000 82e99c10 83992e90 801ba3c4
[ 34.828000] Hi : 00004000
[ 34.828000] Lo : 00000000
[ 34.828000] epc : 8014a890 ioread16+0x4/0xc
[ 34.828000] Tainted: G O
[ 34.828000] ra : 801ba3c4 bcma_host_pci_read16+0x34/0x4c
[ 34.828000] Status: 1100cc03 KERNEL EXL IE
[ 34.828000] Cause : 8080001c
[ 34.828000] PrId : 00019740 (MIPS 74Kc)
[ 34.828000] Modules linked in: nf_nat_irc nf_nat_ftp nf_conntrack_irc
nf_conntrack_ftp ipt_MASQUERADE iptable_nat nf_nat pppoe xt_conntrack
xt_CT xt_NOTRACK iptable_raw xt_state nf_conntrack_ipv4 nf_defrag_ipv4
nf_conntrack pppox ipt_REJECT xt_TCPMSS xt_comment xt_multiport xt_mac
xt_limit iptable_mangle iptable_filter ip_tables xt_tcpudp x_tables
brcmsmac(O) ppp_async ppp_generic slhc brcmutil(O) mac80211(O)
switch_core(O) crc8 crc_ccitt cordic cfg80211(O) compat(O) arc4
aes_generic crypto_blkcipher cryptomgr aead crypto_hash crypto_algapi
[last unloaded: switch_core]
[ 34.828000] Process ifconfig (pid: 853, threadinfo=82e98000,
task=83904d30, tls=77a8e440)
[ 34.828000] Stack : 0000000c 82e07d80 83992e80 801ba2c0 82dfd400
82dfd400 00000000 82e08e68
0000000c 0c8b8482 24189612 6c604830 000305e0 00030007 00000002
00000000
00000000 00000000 00000000 00000000 82e39bf0 82dfd400 00000001
82e39bf0
00000002 82e0de5c 00000001 82e38874 82e99cb0 82e5a894 0000007f
00000001
0003002f 00620007 00050011 0000000c 0c8b8482 24189612 6c604830
82e60000
...
[ 34.828000] Call Trace:
[ 34.828000] [<8014a890>] ioread16+0x4/0xc
[ 34.828000] [<801ba3c4>] bcma_host_pci_read16+0x34/0x4c
[ 34.828000] [<82e08e68>] brcms_c_wme_setparams+0x134/0x1d4 [brcmsmac]
[ 34.828000] [<82e0de5c>] brcms_c_init+0xdac/0xe88 [brcmsmac]
[ 34.828000] [<82e0bfe4>] brcms_c_up+0x35c/0x478 [brcmsmac]
[ 34.828000] [<82e01e00>] brcms_rfkill_set_hw_state+0xec/0x144 [brcmsmac]
[ 34.828000] [<82d910ac>] ieee80211_do_open+0x138/0xaf8 [mac80211]
[ 34.828000] [<801d4a04>] __dev_open+0x144/0x228
[ 34.828000] [<801d4d40>] __dev_change_flags+0xc0/0x160
[ 34.828000] [<801d4e88>] dev_change_flags+0x20/0x6c
[ 34.828000] [<8022df64>] devinet_ioctl+0x2ac/0x7f8
[ 34.828000] [<801bbd80>] sock_ioctl+0x294/0x2f4
[ 34.828000] [<800ab1f4>] do_vfs_ioctl+0x5a4/0x5f8
[ 34.828000] [<800ab298>] sys_ioctl+0x50/0x98
[ 34.828000] [<8000e1d0>] stack_done+0x20/0x40
[ 34.828000]
[ 34.828000]
Code: 03e00008 304200ff 94820000 <03e00008> 3042ffff 94830000
3063ffff 00031200 00031a02
[ 35.104000] ---[ end trace ef9d9b193b014ece ]---
(gdb) l *(brcms_c_wme_setparams+0x134)
0x8e68 is in brcms_c_wme_setparams
(/home/hauke/openwrt/git/build_dir/linux-brcm47xx_uClibc-0.9.33.2/compat-wireless-2012-09-07/drivers/net/wireless/brcm80211/brcmsmac/main.c:4087).
4082 "aifs %d\n", wlc->pub->unit, acp_shm.aifs);
4083 } else {
4084 acp_shm.cwmin = params->cw_min;
4085 acp_shm.cwmax = params->cw_max;
4086 acp_shm.cwcur = acp_shm.cwmin;
4087 acp_shm.bslots =
4088 bcma_read16(wlc->hw->d11core, D11REGOFFS(tsf_random)) &
4089 acp_shm.cwcur;
4090 acp_shm.reggap = acp_shm.bslots + acp_shm.aifs;
4091 /* Indicate the new params to the ucode */
On 12/02/2012 02:36 AM, Hauke Mehrtens wrote:
> Hi Arend,
>
> I am getting a Data bus error when brcmsmac tries to read tsf_random in
> brcms_c_wme_setparams. This does not happen when I add a printk
> somewhere near that read of tsf_random or when I replace the read with a
> constant. In b43 this code works, without this error for me, see
> b43_qos_params_upload(). With the BCM4716 I do not have this problem. It
> could be that this is caused by the broken PCIe controller in the BCM4716.
I can look for known workarounds but I am afraid you know way more about
the BCM4716 than I do.
> I am using brcmsmac from compat-wireless 2012-09-07, but I have also
> tried the most recent brcmsmac code from wireless-testing. brcmsmac uses
> the normal firmware from linux-firmware and I have also tried the new
> firmware posted some days ago with the same result.
>
> I am using a BCM43224 connected to the PCIe controller of the BCM4716.
>
> Are there any special requirements for reading tsf_random under mips?
Not that I recall.
> To make my BCM43224 work I use this patch [0], the "printk("dummy\n");"
> is a hack to workaround the problem I am describing here.
That kind of hacks always make me suspicious about the compiler. I guess
you checked that both brcmsmac and b43 read the same register offset to
get tsf_random, right?
Regards,
Arend
On 12/02/2012 10:47 PM, Nathan Hintz wrote:
> I encountered a similar DBE using the "broadcom-wl" driver (OpenWRT) on a
> Linksys E3000; except that it occurred when reading "ifs_ctl" (offset 0x688).
> I too tried putting in some printk's (in linux_osl.c), which caused the
> problem to go away. I eventually found that doing a dummy read of 4 bytes
> at that particular address just prior to the actual two byte read caused
> the problem to go away as well (an ugly hack). I can reproduce it if you
> want a stack trace.
Does an ndelay(xx), with xx of 10 or so, be as effective as the dummy read? It
is not a lot less of a hack, but it would indicate the source of the problem.
Larry
Hauke Mehrtens <hauke@...> writes:
>
> Hi Arend,
>
> I am getting a Data bus error when brcmsmac tries to read tsf_random in
> brcms_c_wme_setparams. This does not happen when I add a printk
> somewhere near that read of tsf_random or when I replace the read with a
> constant. In b43 this code works, without this error for me, see
> b43_qos_params_upload(). With the BCM4716 I do not have this problem. It
> could be that this is caused by the broken PCIe controller in the BCM4716.
>
> I am using brcmsmac from compat-wireless 2012-09-07, but I have also
> tried the most recent brcmsmac code from wireless-testing. brcmsmac uses
> the normal firmware from linux-firmware and I have also tried the new
> firmware posted some days ago with the same result.
>
> I am using a BCM43224 connected to the PCIe controller of the BCM4716.
>
> Are there any special requirements for reading tsf_random under mips?
>
> To make my BCM43224 work I use this patch [0], the "printk("dummy\n");"
> is a hack to workaround the problem I am describing here.
>
> Hauke
>
> [0]:
>
<snip>
I encountered a similar DBE using the "broadcom-wl" driver (OpenWRT) on a
Linksys E3000; except that it occurred when reading "ifs_ctl" (offset 0x688).
I too tried putting in some printk's (in linux_osl.c), which caused the
problem to go away. I eventually found that doing a dummy read of 4 bytes
at that particular address just prior to the actual two byte read caused
the problem to go away as well (an ugly hack). I can reproduce it if you
want a stack trace.
Nathan