Re: [PATCH v4] pwrap: mediatek: fix FSM timeout issue

From: AngeloGioacchino Del Regno
Date: Fri May 13 2022 - 04:06:47 EST


Il 13/05/22 05:43, Zhiyong Tao ha scritto:
From: "Zhiyong.Tao" <zhiyong.tao@xxxxxxxxxxxx>

Fix pwrap FSM timeout issue which leads the system crash on GFX VSRAM
power on.
The crash log:
[ 3986.543401] mediatek-drm-dp 1c500000.edp_tx: drm_helper_hpd_irq_event
[ 3986.670756] vsram_others: is_enabled() failed: -ETIMEDOUT
[ 3986.670765] mali 13000000.mali: Power on reg 1 failed error = -110
[ 3986.670768] ------------[ cut here ]------------
[ 3986.670770] unbalanced disables for vsram_others
[ 3986.670783] WARNING: CPU: 7 PID: 4125 at drivers/regulator/core.c:2761 _regulator_disable+0x194/0x1a0
[ 3986.670785] Modules linked in: rfcomm algif_hash algif_skcipher af_alg veth uinput btusb btmtk btintel btbcm btrtl xt_cgroup bluetooth uvcvideo videobuf2_vmalloc ecdh_generic ecc mtk_vcodec_dec mtk_vcodec_enc mtk_mdp3 v4l2_h264 mtk_vcodec_common videobuf2_dma_contig mtk_vpu videobuf2_memops v4l2_mem2mem xt_MASQUERADE videobuf2_v4l2 videobuf2_common cros_ec_rpmsg mtk_scp mtk_rpmsg rpmsg_core mtk_scp_ipi ip6table_nat fuse 8021q iio_trig_sysfs cros_ec_sensors cros_ec_lid_angle cros_ec_sensors_core industrialio_triggered_buffer kfifo_buf cros_ec_sensorhub mt7921e mt7921_common mt76_connac_lib lzo_rle mt76 lzo_compress mac80211 cfg80211 zram r8152 mii joydev
[ 3986.670830] CPU: 7 PID: 4125 Comm: mali-cmar-backe Not tainted 5.10.78-CL2781499-v287 #1 b899b40a63da40d4767c6c0e96b6700d2f3eb242
[ 3986.670832] Hardware name: MediaTek Tomato board (DT)
[ 3986.670835] pstate: 60400009 (nZCv daif +PAN -UAO -TCO BTYPE=--)
[ 3986.670838] pc : _regulator_disable+0x194/0x1a0
[ 3986.670840] lr : _regulator_disable+0x194/0x1a0
[ 3986.670842] sp : ffffffc016203a10
[ 3986.670843] x29: ffffffc016203a10 x28: ffffffb7c3186b28
[ 3986.670846] x27: 0000000000000002 x26: fffffffffffffdc8
[ 3986.670848] x25: ffffffc017225000 x24: ffffffb7c0e94880
[ 3986.670851] x23: ffffffb7c31840f0 x22: ffffffd6b4f3e275
[ 3986.670853] x21: ffffffb7c3181a00 x20: ffffffb7c27e7800
[ 3986.670855] x19: ffffffb7c27e7800 x18: 00000000ffff0a10
[ 3986.670857] x17: 0000000000000020 x16: 00000000000000ec
[ 3986.670860] x15: ffffffd6b44fa17c x14: 0000000000000003
[ 3986.670862] x13: 0000000000000004 x12: 0000000000fd8318
[ 3986.670864] x11: c000000100029ccd x10: 00000000ffffffff
[ 3986.670866] x9 : 7dd6d080afd6f400 x8 : 7dd6d080afd6f400
[ 3986.670868] x7 : 0000000000000000 x6 : ffffffd6b5459f0c
[ 3986.670871] x5 : ffffffc016203a58 x4 : 0000000000000000
[ 3986.670873] x3 : ffffffc016203668 x2 : ffffffc016203670
[ 3986.670875] x1 : 0000000100029ccd x0 : 0000000000000024
[ 3986.670878] Call trace:
[ 3986.670880] _regulator_disable+0x194/0x1a0
[ 3986.670883] regulator_disable+0x4c/0x8c

Add a usleep delay to avoid busy read for the H/W status.
If (time_after()) be turn first, it maybe cause the system behavior
crash problem like above. so we change it after sleep delay.

Fixes: 1f022d84bd19 ("soc: mediatek: Add PMIC wrapper for MT8135 and
MT8173 SoCs")

Signed-off-by: Zhiyong.Tao <zhiyong.tao@xxxxxxxxxxxx>
Reviewed-by: Rex-BC Chen <rex-bc.chen@xxxxxxxxxxxx>

Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@xxxxxxxxxxxxx>