* Re: [PATCH] pinctrl: imx: make sure that maps are fully initialized
2018-11-10 17:37 [PATCH] pinctrl: imx: make sure that maps are fully initialized Martin Kaiser
@ 2018-11-10 17:49 ` Fabio Estevam
2018-11-11 22:37 ` Leonard Crestez
2018-11-12 15:04 ` A.s. Dong
2 siblings, 0 replies; 6+ messages in thread
From: Fabio Estevam @ 2018-11-10 17:49 UTC (permalink / raw)
To: Martin Kaiser
Cc: Dong Aisheng, Shawn Guo, Stefan Agner, Sascha Hauer,
Linus Walleij, open list:GPIO SUBSYSTEM, linux-kernel
Hi Martin,
On Sat, Nov 10, 2018 at 3:39 PM Martin Kaiser <martin@kaiser.cx> wrote:
>
> The commit that added scu based pinctrl support introduced a regression
> for the mmio case. In the for-loop where the maps are initialized, we
> end up creating a partially initialized map in some cases. This causes a
> kernel panic when such a map is used at a later stage.
>
> When scu is not used and and a pin uses the default pad configuration,
> the current code sets the map's type and data.configs.group_or_pin but
> no configs or num_configs are set. The previous code without scu
> support did not create a new map at all in this case and did not
> increment the counter.
>
> Fix this by setting the map type and data.configs.group_or_pin only when
> a configuration was set before, either by scu or by mmio using a
> non-default pad config. Increment the counter only if a new map was
> created.
>
> Fixes: b96eea718bf6 ("pinctrl: fsl: add scu based pinctrl support")
> Cc: A.s. Dong <aisheng.dong@nxp.com>
> Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Thanks for the fix:
Reviewed-by: Fabio Estevam <festevam@gmail.com>
^ permalink raw reply [flat|nested] 6+ messages in thread* Re: [PATCH] pinctrl: imx: make sure that maps are fully initialized
2018-11-10 17:37 [PATCH] pinctrl: imx: make sure that maps are fully initialized Martin Kaiser
2018-11-10 17:49 ` Fabio Estevam
@ 2018-11-11 22:37 ` Leonard Crestez
2018-11-12 15:37 ` Martin Kaiser
2018-11-12 15:04 ` A.s. Dong
2 siblings, 1 reply; 6+ messages in thread
From: Leonard Crestez @ 2018-11-11 22:37 UTC (permalink / raw)
To: Martin Kaiser, A.s. Dong, Fabio Estevam, Linus Walleij
Cc: Shawn Guo, Stefan Agner, Pengutronix Kernel Team, linux-gpio,
linux-kernel, dl-linux-imx
On 11/10/18 7:37 PM, Martin Kaiser wrote:
> The commit that added scu based pinctrl support introduced a regression
> for the mmio case. In the for-loop where the maps are initialized, we
> end up creating a partially initialized map in some cases. This causes a
> kernel panic when such a map is used at a later stage.
>
> Fixes: b96eea718bf6 ("pinctrl: fsl: add scu based pinctrl support")
> Cc: A.s. Dong <aisheng.dong@nxp.com>
> diff --git a/drivers/pinctrl/freescale/pinctrl-imx.c b/drivers/pinctrl/freescale/pinctrl-imx.c
> @@ -108,9 +108,6 @@ static int imx_dt_node_to_map(struct pinctrl_dev *pctldev,
> new_map++;
> for (i = j = 0; i < grp->num_pins; i++) {
> pin = &((struct imx_pin *)(grp->data))[i];
> - new_map[j].type = PIN_MAP_TYPE_CONFIGS_PIN;
> - new_map[j].data.configs.group_or_pin =
> - pin_get_name(pctldev, pin->pin);
>
> if (info->flags & IMX_USE_SCU) {
> /*
> @@ -126,7 +123,12 @@ static int imx_dt_node_to_map(struct pinctrl_dev *pctldev,
> new_map[j].data.configs.num_configs = 1;
> }
>
> - j++;
> + if (new_map[j].data.configs.num_configs) {
> + new_map[j].type = PIN_MAP_TYPE_CONFIGS_PIN;
> + new_map[j].data.configs.group_or_pin =
> + pin_get_name(pctldev, pin->pin);
> + j++;
> + }
Sorry but I don't think this is correct.
The new_map array is allocated with kmalloc_array so we can't rely on
new_map[j].data.configs.num_configs being initialized to zero unless
assigned to.
I was looking at the ssue before I saw your patch and posted an
different fix. I can't find a give link right now, it takes a while to
show up in public archives.
^ permalink raw reply [flat|nested] 6+ messages in thread* RE: [PATCH] pinctrl: imx: make sure that maps are fully initialized
2018-11-10 17:37 [PATCH] pinctrl: imx: make sure that maps are fully initialized Martin Kaiser
2018-11-10 17:49 ` Fabio Estevam
2018-11-11 22:37 ` Leonard Crestez
@ 2018-11-12 15:04 ` A.s. Dong
2018-11-12 15:09 ` Fabio Estevam
2 siblings, 1 reply; 6+ messages in thread
From: A.s. Dong @ 2018-11-12 15:04 UTC (permalink / raw)
To: Martin Kaiser, Fabio Estevam, Shawn Guo, Stefan Agner,
Pengutronix Kernel Team, Linus Walleij, linux-gpio, linux-kernel,
Leonard Crestez
Hi Martin,
> -----Original Message-----
> From: Martin Kaiser [mailto:martin@kaiser.cx]
> Sent: Sunday, November 11, 2018 1:37 AM
[...]
>
> The commit that added scu based pinctrl support introduced a regression for
> the mmio case. In the for-loop where the maps are initialized, we end up
> creating a partially initialized map in some cases. This causes a kernel panic
> when such a map is used at a later stage.
>
> When scu is not used and and a pin uses the default pad configuration, the
> current code sets the map's type and data.configs.group_or_pin but no configs
> or num_configs are set. The previous code without scu support did not create a
> new map at all in this case and did not increment the counter.
>
> Fix this by setting the map type and data.configs.group_or_pin only when a
> configuration was set before, either by scu or by mmio using a non-default pad
> config. Increment the counter only if a new map was created.
>
> Fixes: b96eea718bf6 ("pinctrl: fsl: add scu based pinctrl support")
> Cc: A.s. Dong <aisheng.dong@nxp.com>
> Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Thanks for reporting this issue.
The original code intended to save 3 common lines for both SCU and MMIO pads,
but it did change the code logic a bit which resulted in a regression specificly
for NO_PAD_CTL cases used on a few legacy boards. (NO_PAD_CTL is deprecated
for new boards dts).
I agree with Leonard that this fixes is not so good.
A simpler and proper fix could be:
diff --git a/drivers/pinctrl/freescale/pinctrl-imx.c b/drivers/pinctrl/freescale/pinctrl-imx.c
index b704f7c..ab962f5 100644
--- a/drivers/pinctrl/freescale/pinctrl-imx.c
+++ b/drivers/pinctrl/freescale/pinctrl-imx.c
@@ -109,6 +109,15 @@ static int imx_dt_node_to_map(struct pinctrl_dev *pctldev,
new_map++;
for (i = j = 0; i < grp->num_pins; i++) {
pin = &((struct imx_pin *)(grp->data))[i];
+
+ /*
+ * We only create config maps for SCU pads or MMIO pads that
+ * are not using the default config (a.k.a NO_PAD_CTL)
+ */
+ if (!(info->flags & IMX_USE_SCU) &&
+ (pin->conf.mmio.config & IMX_NO_PAD_CTL))
+ continue;
+
new_map[j].type = PIN_MAP_TYPE_CONFIGS_PIN;
new_map[j].data.configs.group_or_pin =
pin_get_name(pctldev, pin->pin);
@@ -121,7 +130,7 @@ static int imx_dt_node_to_map(struct pinctrl_dev *pctldev,
new_map[j].data.configs.configs =
(unsigned long *)&pin->conf.scu;
new_map[j].data.configs.num_configs = 2;
- } else if (!(pin->conf.mmio.config & IMX_NO_PAD_CTL)) {
+ } else {
new_map[j].data.configs.configs =
&pin->conf.mmio.config;
new_map[j].data.configs.num_configs = 1;
@@ -550,6 +559,8 @@ static void imx_pinctrl_parse_pin_mmio(struct imx_pinctrl *ipctl,
pin_mmio->config = config & ~IMX_PAD_SION;
}
+ *list_p = list; //thanks to Leonard for pointing out this
+
dev_dbg(ipctl->dev, "%s: 0x%x 0x%08lx", info->pins[*pin_id].name,
pin_mmio->mux_mode, pin_mmio->config);
As you both had a patch, i'm not sure who i'm going to expect for a updated one,
so I would send it my own with both of your tags.
Thanks for the effort.
Regards
Dong Aisheng
> ---
>
> A crash looks like this
>
> [ 4.028525] Unable to handle kernel NULL pointer dereference at virtual
> address 00000000
> [ 4.036667] pgd = (ptrval)
> [ 4.039505] [00000000] *pgd=00000000
> [ 4.043162] Internal error: Oops: 5 [#1] ARM
> [ 4.047466] Modules linked in:
> [ 4.050582] CPU: 0 PID: 1 Comm: swapper Not tainted
> 4.20.0-rc1-next-20181106+ #2269
> [ 4.058270] Hardware name: Freescale i.MX25 (Device Tree Support)
> [ 4.064456] PC is at strcmp+0xc/0x40
> [ 4.068111] LR is at create_pinctrl+0x104/0x36c
> [ 4.072680] pc : [<c05699d8>] lr : [<c0220bc8>] psr: 60000013
> [ 4.078980] sp : d38bbce8 ip : d38bbcf8 fp : d38bbcf4
> [ 4.084236] r10: d3b2e0e0 r9 : d3b2e0c0 r8 : 00000000
> [ 4.089494] r7 : d3b2e238 r6 : d3b2e1c0 r5 : 00000000 r4 :
> d3b0df00
> [ 4.096054] r3 : 85981bcd r2 : 85981bcd r1 : d3a729e0 r0 :
> 00000000
> [ 4.102615] Flags: nZCv IRQs on FIQs on Mode SVC_32 ISA ARM
> Segment none
> [ 4.109786] Control: 0005317f Table: 80004000 DAC: 00000053
> [ 4.115569] Process swapper (pid: 1, stack limit = 0x(ptrval))
> [ 4.121437] Stack: (0xd38bbce8 to 0xd38bc000)
> [ 4.125849] bce0: d38bbd3c d38bbcf8 c0220bc8
> c05699dc d38bbd3c c06adb58
> [ 4.134084] bd00: c078cc54 d3b0df0c d3a729e0 00000002 00000000
> c078cc40 d3b0e800 d3b0e800
> [ 4.142318] bd20: c07c7ba0 c07e8ed0 c07c0480 c07cf820 d38bbd54
> d38bbd40 c0220eec c0220ad4
> [ 4.150553] bd40: d3b0e800 d3b2e230 d38bbd74 d38bbd58 c0220f44
> c0220e40 c028e24c d3b0e800
> [ 4.158787] bd60: 00000000 d3b0df50 d38bbd94 d38bbd78 c02a6e38
> c0220f18 d3b0e948 d3b0e800
> [ 4.167019] bd80: 00000000 c07a66d8 d38bbdd4 d38bbd98 c028a614
> c02a6dfc 20000013 85981bcd
> [ 4.175254] bda0: c07a66d8 00000000 d3b0e800 d3b0e800 c07a66d8
> c07a66d8 c028acac c07c7ac8
> [ 4.183489] bdc0: c07c0480 c07cf820 d38bbe0c d38bbdd8 c028ac54
> c028a4fc c03a8acc c03a69f0
> [ 4.191722] bde0: d38bbe0c d38bbdf0 d3b0e800 d3b0e834 c07a66d8
> c028acac c07c7ac8 c07c0480
> [ 4.199954] be00: d38bbe2c d38bbe10 c028ad94 c028aaf0 00000000
> d3b0e800 c075b028 c07a66d8
> [ 4.208189] be20: d38bbe5c d38bbe30 c028867c c028acbc d38bbe68
> d38fe64c d3b08f30 85981bcd
> [ 4.216421] be40: c07a66d8 00000000 d3b29980 c0798384 d38bbe6c
> d38bbe60 c028a01c c0288618
> [ 4.224655] be60: d38bbe9c d38bbe70 c0289954 c028a004 c069a541
> c07176a8 d38bbe9c c07a66d8
> [ 4.232890] be80: 00000006 c07176a8 c075b028 00000000 d38bbeb4
> d38bbea0 c028b99c c028983c
> [ 4.241124] bea0: c075687c 00000006 d38bbec4 d38bbeb8 c02e9150
> c028b8ec d38bbed4 d38bbec8
> [ 4.249357] bec0: c07176c8 c02e9108 d38bbf4c d38bbed8 c000ab54
> c07176b8 c0034038 c06ed774
> [ 4.257589] bee0: 00000000 d38bbe00 d38bbee4 c06eb8ec 00000085
> c06eb8ec 00000000 00000006
> [ 4.265822] bf00: 00000006 00000086 c06eae74 c06ed764 d38bbf34
> d3ffcbbc d3ffcbc6 85981bcd
> [ 4.274057] bf20: 00000000 85981bcd c075687c 00000006 c07c0480
> c0733830 00000086 c07cf820
> [ 4.282289] bf40: d38bbf94 d38bbf50 c06ee0e8 c000aadc 00000006
> 00000006 00000000 c06ed764
> [ 4.290520] bf60: 00000000 c0023240 d38bbf84 00000000 c0571424
> 00000000 00000000 00000000
> [ 4.298753] bf80: 00000000 00000000 d38bbfac d38bbf98 c057143c
> c06edf5c 60000053 00000000
> [ 4.306982] bfa0: 00000000 d38bbfb0 c00090e0 c0571434 00000000
> 00000000 00000000 00000000
> [ 4.315210] bfc0: 00000000 00000000 00000000 00000000 00000000
> 00000000 00000000 00000000
> [ 4.323436] bfe0: 00000000 00000000 00000000 00000000 00000013
> 00000000 00000000 00000000
> [ 4.331634] Backtrace:
> [ 4.334199] [<c05699cc>] (strcmp) from [<c0220bc8>]
> (create_pinctrl+0x104/0x36c)
> [ 4.341679] [<c0220ac4>] (create_pinctrl) from [<c0220eec>]
> (pinctrl_get+0xbc/0xd8)
> [ 4.349403] r10:c07cf820 r9:c07c0480 r8:c07e8ed0 r7:c07c7ba0
> r6:d3b0e800 r5:d3b0e800
> [ 4.357265] r4:c078cc40
> [ 4.359869] [<c0220e30>] (pinctrl_get) from [<c0220f44>]
> (devm_pinctrl_get+0x3c/0x7c)
> [ 4.367745] r5:d3b2e230 r4:d3b0e800
> [ 4.371416] [<c0220f08>] (devm_pinctrl_get) from [<c02a6e38>]
> (pinctrl_bind_pins+0x4c/0x258)
> [ 4.379907] r6:d3b0df50 r5:00000000 r4:d3b0e800 r3:c028e24c
> [ 4.385654] [<c02a6dec>] (pinctrl_bind_pins) from [<c028a614>]
> (really_probe+0x128/0x40c)
> [ 4.393883] r6:c07a66d8 r5:00000000 r4:d3b0e800 r3:d3b0e948
> [ 4.399613] [<c028a4ec>] (really_probe) from [<c028ac54>]
> (driver_probe_device+0x174/0x1cc)
> [ 4.408031] r10:c07cf820 r9:c07c0480 r8:c07c7ac8 r7:c028acac
> r6:c07a66d8 r5:c07a66d8
> [ 4.415893] r4:d3b0e800
> [ 4.418497] [<c028aae0>] (driver_probe_device) from [<c028ad94>]
> (__driver_attach+0xe8/0x138)
> [ 4.427084] r9:c07c0480 r8:c07c7ac8 r7:c028acac r6:c07a66d8
> r5:d3b0e834 r4:d3b0e800
>
>
> I'm not sure what exactly is required to trigger the problem. I saw the crash for
> a dt device node like this where pinctrl links to two nodes with the actual pin
> settings.
>
> &spi1 {
> pinctrl-names = "default";
> pinctrl-0 = <&pinctrl_spi1_common &pinctrl_spi1_cs>; }
>
> pinctrl_spi1_common contains a couple of pins with non-default pad settings
> and pinctrl_spi1_cs contains a single pin that uses the default pad
> configuration.
>
> pinctrl_spi1_cs: spi1_cs_grp {
> fsl,pins = <
> MX25_PAD_CSPI1_SS0__GPIO_1_16 0x80000000
> >;
> };
>
>
> drivers/pinctrl/freescale/pinctrl-imx.c | 10 ++++++----
> 1 file changed, 6 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/pinctrl/freescale/pinctrl-imx.c
> b/drivers/pinctrl/freescale/pinctrl-imx.c
> index 78d33dfb4d2d..0c4ad8d19226 100644
> --- a/drivers/pinctrl/freescale/pinctrl-imx.c
> +++ b/drivers/pinctrl/freescale/pinctrl-imx.c
> @@ -108,9 +108,6 @@ static int imx_dt_node_to_map(struct pinctrl_dev
> *pctldev,
> new_map++;
> for (i = j = 0; i < grp->num_pins; i++) {
> pin = &((struct imx_pin *)(grp->data))[i];
> - new_map[j].type = PIN_MAP_TYPE_CONFIGS_PIN;
> - new_map[j].data.configs.group_or_pin =
> - pin_get_name(pctldev, pin->pin);
>
> if (info->flags & IMX_USE_SCU) {
> /*
> @@ -126,7 +123,12 @@ static int imx_dt_node_to_map(struct pinctrl_dev
> *pctldev,
> new_map[j].data.configs.num_configs = 1;
> }
>
> - j++;
> + if (new_map[j].data.configs.num_configs) {
> + new_map[j].type = PIN_MAP_TYPE_CONFIGS_PIN;
> + new_map[j].data.configs.group_or_pin =
> + pin_get_name(pctldev, pin->pin);
> + j++;
> + }
> }
>
> dev_dbg(pctldev->dev, "maps: function %s group %s num %d\n",
> --
> 2.1.4
^ permalink raw reply related [flat|nested] 6+ messages in thread