1 From: Sergey Ryazanov <ryazanov.s.a@gmail.com>
2 Date: Wed, 4 Mar 2015 05:12:10 +0300
3 Subject: [PATCH] ath5k: channel change fix
5 ath5k updates the channel pointer and after that it stops the Rx logic
6 and apply channel to HW. In case of channel switch, such sequence
7 creates a small window when a frame, which is received on the old
8 channel is considered as a frame received on the new one.
10 The most notable consequence of this situation occurs during the switch
11 from 2 GHz band (CCK+OFDM) to the 5GHz band (OFDM-only). Frame received
12 with CCK rate, e.g. beacon received at the 1mbps, causes the following
15 WARNING: at ath5k/base.c:589 ath5k_tasklet_rx+0x318/0x6ec [ath5k]()
19 [<802656a8>] show_stack+0x48/0x70
20 [<802dd92c>] warn_slowpath_common+0x88/0xbc
21 [<802dd98c>] warn_slowpath_fmt+0x2c/0x38
22 [<81b51be8>] ath5k_tasklet_rx+0x318/0x6ec [ath5k]
23 [<8028ac64>] tasklet_action+0x8c/0xf0
24 [<80075804>] __do_softirq+0x180/0x32c
25 [<80196ce8>] irq_exit+0x54/0x70
26 [<80041848>] ret_from_irq+0x0/0x4
27 [<80182fdc>] ioread32+0x4/0xc
28 [<81b4c42c>] ath5k_hw_set_sleep_clock+0x2ec/0x474 [ath5k]
29 [<81b4cf28>] ath5k_hw_reset+0x50/0xeb8 [ath5k]
30 [<81b50900>] ath5k_reset+0xd4/0x310 [ath5k]
31 [<81b557e8>] ath5k_config+0x4c/0x104 [ath5k]
32 [<80d01770>] ieee80211_hw_config+0x2f4/0x35c [mac80211]
33 [<80d09aa8>] ieee80211_scan_work+0x2e4/0x414 [mac80211]
34 [<8022c3f4>] process_one_work+0x28c/0x400
35 [<802df8f8>] worker_thread+0x258/0x3c0
36 [<801b5710>] kthread+0xe0/0xec
37 [<800418a8>] ret_from_kernel_thread+0x14/0x1c
39 The easiest way to reproduce this warning is to run scan with dualband
40 NIC in noisy environments, when the channel 11 runs multiple APs. In my
41 tests if the APs num >= 12, the warning appears in the first few
44 In order to fix this, the Rx disable code moved to a higher level and
45 placed before the channel pointer update. This is also makes the code a
46 bit more symmetrical, since we disable and enable the Rx in the same
49 In fact, at the pointer update time new frames should not appear,
50 because interrupt generation at this point should already be disabled.
51 The next patch should address this issue.
53 CC: Jiri Slaby <jirislaby@gmail.com>
54 CC: Nick Kossifidis <mickflemm@gmail.com>
55 CC: Luis R. Rodriguez <mcgrof@do-not-panic.com>
56 Reported-by: Christophe Prevotaux <cprevotaux@nltinc.com>
57 Tested-by: Christophe Prevotaux <cprevotaux@nltinc.com>
58 Tested-by: Eric Bree <ebree@nltinc.com>
59 Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
62 --- a/drivers/net/wireless/ath/ath5k/base.c
63 +++ b/drivers/net/wireless/ath/ath5k/base.c
64 @@ -2858,7 +2858,7 @@ ath5k_reset(struct ath5k_hw *ah, struct
66 struct ath_common *common = ath5k_hw_common(ah);
69 + bool fast = chan && modparam_fastchanswitch ? 1 : 0;
71 ATH5K_DBG(ah, ATH5K_DEBUG_RESET, "resetting\n");
73 @@ -2876,11 +2876,29 @@ ath5k_reset(struct ath5k_hw *ah, struct
74 * so we should also free any remaining
76 ath5k_drain_tx_buffs(ah);
79 + ath5k_hw_stop_rx_pcu(ah);
83 + * Note: If DMA didn't stop continue
84 + * since only a reset will fix it.
86 + ret = ath5k_hw_dma_stop(ah);
88 + /* RF Bus grant won't work if we have pending
92 + ATH5K_DBG(ah, ATH5K_DEBUG_RESET,
93 + "DMA didn't stop, falling back to normal reset\n");
100 - fast = ((chan != NULL) && modparam_fastchanswitch) ? 1 : 0;
102 ret = ath5k_hw_reset(ah, ah->opmode, ah->curchan, fast, skip_pcu);
104 ATH5K_ERR(ah, "can't reset hardware (%d)\n", ret);
105 --- a/drivers/net/wireless/ath/ath5k/reset.c
106 +++ b/drivers/net/wireless/ath/ath5k/reset.c
107 @@ -1169,30 +1169,6 @@ ath5k_hw_reset(struct ath5k_hw *ah, enum
108 if (ah->ah_version == AR5K_AR5212)
109 ath5k_hw_set_sleep_clock(ah, false);
114 - ath5k_hw_stop_rx_pcu(ah);
119 - * Note: If DMA didn't stop continue
120 - * since only a reset will fix it.
122 - ret = ath5k_hw_dma_stop(ah);
124 - /* RF Bus grant won't work if we have pending
127 - ATH5K_DBG(ah, ATH5K_DEBUG_RESET,
128 - "DMA didn't stop, falling back to normal reset\n");
130 - /* Non fatal, just continue with
135 mode = channel->hw_value;