On apq8016, apq8084 and apq8074 platforms, when we want to do a
software reset, we need to poke some additional vendor specific
registers. If we don't do so, the following error message appears:
mmc0: Reset 0x1 never completed.
sdhci: =========== REGISTER DUMP (mmc0)===========
sdhci: Sys addr: 0x00000000 | Version: 0x00002e02
sdhci: Blk size: 0x00004000 | Blk cnt: 0x00000000
sdhci: Argument: 0x00000000 | Trn mode: 0x00000000
sdhci: Present: 0x01f80000 | Host ctl: 0x00000000
sdhci: Power: 0x00000000 | Blk gap: 0x00000000
sdhci: Wake-up: 0x00000000 | Clock: 0x00000003
sdhci: Timeout: 0x00000000 | Int stat: 0x00000000
sdhci: Int enab: 0x00000000 | Sig enab: 0x00000000
sdhci: AC12 err: 0x00000000 | Slot int: 0x00000000
sdhci: Caps: 0x322dc8b2 | Caps_1: 0x00008007
sdhci: Cmd: 0x00000000 | Max curr: 0x00000000
sdhci: Host ctl2: 0x00000000
sdhci: ADMA Err: 0x00000000 | ADMA Ptr: 0x0000000000000000
sdhci: ===========================================
Fix it by implementing the custom sdhci_reset() function,
which does what is needed.
Signed-off-by: Georgi Djakov <[email protected]>
---
drivers/mmc/host/sdhci-msm.c | 19 ++++++++++++++++++-
1 file changed, 18 insertions(+), 1 deletion(-)
diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c
index 8ef44a2a2fd9..87a124a37408 100644
--- a/drivers/mmc/host/sdhci-msm.c
+++ b/drivers/mmc/host/sdhci-msm.c
@@ -505,6 +505,23 @@ static irqreturn_t sdhci_msm_pwr_irq(int irq, void *data)
return IRQ_HANDLED;
}
+void sdhci_msm_reset(struct sdhci_host *host, u8 mask)
+{
+ struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
+ struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
+
+ if (mask & SDHCI_RESET_ALL) {
+ u32 val = readl_relaxed(msm_host->core_mem + CORE_POWER);
+
+ val |= CORE_SW_RST;
+ writel_relaxed(val, msm_host->core_mem + CORE_POWER);
+
+ sdhci_msm_voltage_switch(host);
+ }
+
+ sdhci_reset(host, mask);
+}
+
static const struct of_device_id sdhci_msm_dt_match[] = {
{ .compatible = "qcom,sdhci-msm-v4" },
{},
@@ -514,7 +531,7 @@ MODULE_DEVICE_TABLE(of, sdhci_msm_dt_match);
static const struct sdhci_ops sdhci_msm_ops = {
.platform_execute_tuning = sdhci_msm_execute_tuning,
- .reset = sdhci_reset,
+ .reset = sdhci_msm_reset,
.set_clock = sdhci_set_clock,
.set_bus_width = sdhci_set_bus_width,
.set_uhs_signaling = sdhci_msm_set_uhs_signaling,
Hi Georgi,
I collected some info on this problem. May be below info might help you.
I think "Reset 0x1" problem is occurring because of below call stack.
SDHCI_RESET_ALL to SDHCI_SOFTWARE_RESET register will anyway trigger the
sdhci_msm_pwr_irq.
But I think the problem is that the above occurs in spinlock context
and because of only one core the IRQ will never be serviced, hence you
were seeing (Reset 0x1) error.
[ 12.583245] systemd-journald[1236]: Received SIGTERM from PID 1
(systemd-shutdow).
[ 12.673684] EXT4-fs (mmcblk0p10): re-mounted. Opts: data=ordered
[ 12.678224] EXT4-fs (mmcblk0p10): re-mounted. Opts: data=ordered
[ 13.698330] mmc0: Reset 0x1 never completed.
[ 13.698353] sdhci: =========== REGISTER DUMP (mmc0)===========
[ 13.701659] sdhci: Sys addr: 0x00000000 | Version: 0x00002e02
[ 13.707301] sdhci: Blk size: 0x00004000 | Blk cnt: 0x00000000
[ 13.713117] sdhci: Argument: 0x00000000 | Trn mode: 0x00000000
[ 13.718933] sdhci: Present: 0x01f80000 | Host ctl: 0x00000000
[ 13.724750] sdhci: Power: 0x00000000 | Blk gap: 0x00000000
[ 13.730564] sdhci: Wake-up: 0x00000000 | Clock: 0x00000003
[ 13.736381] sdhci: Timeout: 0x00000000 | Int stat: 0x00000000
[ 13.742197] sdhci: Int enab: 0x00000000 | Sig enab: 0x00000000
[ 13.748012] sdhci: AC12 err: 0x00000000 | Slot int: 0x00000000
[ 13.753830] sdhci: Caps: 0x322dc8b2 | Caps_1: 0x00008007
[ 13.759644] sdhci: Cmd: 0x00000000 | Max curr: 0x00000000
[ 13.765461] sdhci: Host ctl2: 0x00000000
[ 13.771275] sdhci: ADMA Err: 0x00000000 | ADMA Ptr: 0x0000000000000000
[ 13.775357] sdhci: ===========================================
[ 13.781698] CPU: 0 PID: 1 Comm: systemd-shutdow Not tainted
4.9.0-rc5-00222-g59ac3c0-dirty #9
[ 13.787514] Hardware name: Qualcomm Technologies, Inc. APQ 8016 SBC (DT)
[ 13.796104] Call trace:
[ 13.802878] [<ffff0000080882c4>] dump_backtrace+0x0/0x1a8
[ 13.805049] [<ffff000008088480>] show_stack+0x14/0x1c
[ 13.810602] [<ffff000008366274>] dump_stack+0x8c/0xb0
[ 13.815640] [<ffff00000870eab0>] sdhci_reset+0xd8/0x114
[ 13.820670] [<ffff00000870ee30>] sdhci_do_reset+0x48/0x7c
[ 13.825706] [<ffff00000870ef20>] sdhci_init+0xbc/0x110
[ 13.831260] [<ffff000008710c44>] sdhci_set_ios+0x68/0x59c
[ 13.836299] [<ffff0000086f95f4>] mmc_set_initial_state+0xc0/0xcc
[ 13.841767] [<ffff0000086f983c>] mmc_power_off.part.22+0x28/0x40
[ 13.847841] [<ffff0000086f9b5c>] mmc_power_off+0x14/0x1c
[ 13.853832] [<ffff0000086fc754>] _mmc_suspend+0x1e4/0x260
[ 13.859127] [<ffff0000086fe104>] mmc_shutdown+0x2c/0x60
[ 13.864421] [<ffff0000086faa00>] mmc_bus_shutdown+0x40/0x74
[ 13.869458] [<ffff0000084f9280>] device_shutdown+0xf0/0x1a8
[ 13.875013] [<ffff0000080db908>] kernel_restart_prepare+0x34/0x3c
[ 13.880568] [<ffff0000080db9e4>] kernel_restart+0x14/0x74
[ 13.886817] [<ffff0000080dbd2c>] SyS_reboot+0x178/0x244
[ 13.892198] [<ffff000008082ef0>] el0_svc_naked+0x24/0x28
[ 13.897300] mmc0: sdhci_msm_pwr_irq:
[ 13.902799] mmc0: sdhci_msm_voltage_switch: irq_status 9
[ 13.906355] mmc0: sdhci_msm_voltage_switch: irq_status 9, irq_ack 5
To prove above I tried this and the problem goes away. But I still dont
think that the below approach is correct. As it will still trigger a
pwr_irq as well.
diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
index 62aedf1..01e611c 100644
--- a/drivers/mmc/host/sdhci.c
+++ b/drivers/mmc/host/sdhci.c
@@ -174,6 +174,8 @@ void sdhci_reset(struct sdhci_host *host, u8 mask)
/* Reset-all turns off SD Bus Power */
if (host->quirks2 & SDHCI_QUIRK2_CARD_ON_NEEDS_BUS_ON)
sdhci_runtime_pm_bus_off(host);
+ if (host->ops->voltage_switch)
+ host->ops->voltage_switch(host);
}
/* Wait max 100 ms */
On 11/22/2016 9:20 PM, Georgi Djakov wrote:
> On apq8016, apq8084 and apq8074 platforms, when we want to do a
> software reset, we need to poke some additional vendor specific
> registers. If we don't do so, the following error message appears:
>
> mmc0: Reset 0x1 never completed.
> sdhci: =========== REGISTER DUMP (mmc0)===========
> sdhci: Sys addr: 0x00000000 | Version: 0x00002e02
> sdhci: Blk size: 0x00004000 | Blk cnt: 0x00000000
> sdhci: Argument: 0x00000000 | Trn mode: 0x00000000
> sdhci: Present: 0x01f80000 | Host ctl: 0x00000000
> sdhci: Power: 0x00000000 | Blk gap: 0x00000000
> sdhci: Wake-up: 0x00000000 | Clock: 0x00000003
> sdhci: Timeout: 0x00000000 | Int stat: 0x00000000
> sdhci: Int enab: 0x00000000 | Sig enab: 0x00000000
> sdhci: AC12 err: 0x00000000 | Slot int: 0x00000000
> sdhci: Caps: 0x322dc8b2 | Caps_1: 0x00008007
> sdhci: Cmd: 0x00000000 | Max curr: 0x00000000
> sdhci: Host ctl2: 0x00000000
> sdhci: ADMA Err: 0x00000000 | ADMA Ptr: 0x0000000000000000
> sdhci: ===========================================
>
> Fix it by implementing the custom sdhci_reset() function,
> which does what is needed.
>
> Signed-off-by: Georgi Djakov <[email protected]>
> ---
> drivers/mmc/host/sdhci-msm.c | 19 ++++++++++++++++++-
> 1 file changed, 18 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/mmc/host/sdhci-msm.c b/drivers/mmc/host/sdhci-msm.c
> index 8ef44a2a2fd9..87a124a37408 100644
> --- a/drivers/mmc/host/sdhci-msm.c
> +++ b/drivers/mmc/host/sdhci-msm.c
> @@ -505,6 +505,23 @@ static irqreturn_t sdhci_msm_pwr_irq(int irq, void *data)
> return IRQ_HANDLED;
> }
>
> +void sdhci_msm_reset(struct sdhci_host *host, u8 mask)
> +{
> + struct sdhci_pltfm_host *pltfm_host = sdhci_priv(host);
> + struct sdhci_msm_host *msm_host = sdhci_pltfm_priv(pltfm_host);
> +
> + if (mask & SDHCI_RESET_ALL) {
> + u32 val = readl_relaxed(msm_host->core_mem + CORE_POWER);
> +
> + val |= CORE_SW_RST;
> + writel_relaxed(val, msm_host->core_mem + CORE_POWER);
Not required as sdhci_reset register should anyway reset and trigger a
pwr_irq. Because you are not servicing that IRQ the problem is present.
> +
> + sdhci_msm_voltage_switch(host);
> + }
> +
> + sdhci_reset(host, mask);
This I think is sufficient.
> +}
> +
> static const struct of_device_id sdhci_msm_dt_match[] = {
> { .compatible = "qcom,sdhci-msm-v4" },
> {},
> @@ -514,7 +531,7 @@ MODULE_DEVICE_TABLE(of, sdhci_msm_dt_match);
>
> static const struct sdhci_ops sdhci_msm_ops = {
> .platform_execute_tuning = sdhci_msm_execute_tuning,
> - .reset = sdhci_reset,
> + .reset = sdhci_msm_reset,
> .set_clock = sdhci_set_clock,
> .set_bus_width = sdhci_set_bus_width,
> .set_uhs_signaling = sdhci_msm_set_uhs_signaling,
> --
> To unsubscribe from this list: send the line "unsubscribe linux-mmc" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>
--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
On 11/24/2016 06:06 PM, Ritesh Harjani wrote:
> Hi Georgi,
>
> I collected some info on this problem. May be below info might help you.
>
> I think "Reset 0x1" problem is occurring because of below call stack.
> SDHCI_RESET_ALL to SDHCI_SOFTWARE_RESET register will anyway trigger the
> sdhci_msm_pwr_irq.
>
> But I think the problem is that the above occurs in spinlock context
> and because of only one core the IRQ will never be serviced, hence you
> were seeing (Reset 0x1) error.
>
Hi Ritesh,
Thanks for looking into this. So yes, its called in spinlock context and
what we need to do is just handle the power irq after writing to reset
register.
[..]
>
> To prove above I tried this and the problem goes away. But I still dont
> think that the below approach is correct. As it will still trigger a
> pwr_irq as well.
>
>
> diff --git a/drivers/mmc/host/sdhci.c b/drivers/mmc/host/sdhci.c
> index 62aedf1..01e611c 100644
> --- a/drivers/mmc/host/sdhci.c
> +++ b/drivers/mmc/host/sdhci.c
> @@ -174,6 +174,8 @@ void sdhci_reset(struct sdhci_host *host, u8 mask)
> /* Reset-all turns off SD Bus Power */
> if (host->quirks2 & SDHCI_QUIRK2_CARD_ON_NEEDS_BUS_ON)
> sdhci_runtime_pm_bus_off(host);
> + if (host->ops->voltage_switch)
> + host->ops->voltage_switch(host);
> }
>
Yes, our own reset() function that additionally handles the irq will work.
Thanks,
Georgi