Hello Qiang Zhao,
Am 15.04.2020 um 10:10 schrieb Qiang Zhao:
On 2020/4/15 12:46 Heiko Schocher hs@denx.de
-----Original Message----- From: Heiko Schocher hs@denx.de Sent: 2020年4月15日 12:46 To: Qiang Zhao qiang.zhao@nxp.com Cc: U-Boot Mailing List u-boot@lists.denx.de; Holger Brunck holger.brunck@ch.abb.com; Mario Six mario.six@gdsys.cc; Wolfgang Denk wd@denx.de; Holger Brunck holger.brunck@ch.abb.com Subject: Re: [RFC PATCH] powerpc, qe: add DTS support for parallel I/O ports
Hello Qiang Zhao,
Am 09.04.2020 um 08:58 schrieb Qiang Zhao:
On 2020/2/18 17:05, Heiko Schocher <hs@denx.de mailto:hs@denx.de>
wrote:
-----Original Message-----
From: Heiko Schocher hs@denx.de
Sent: 2020年2月18日17:05
To: U-Boot Mailing List u-boot@lists.denx.de
Cc: Heiko Schocher hs@denx.de; Holger Brunck
holger.brunck@ch.abb.com; Mario Six mario.six@gdsys.cc; Qiang Zhao
qiang.zhao@nxp.com; Wolfgang Denk wd@denx.de
Subject: [RFC PATCH] powerpc, qe: add DTS support for parallel I/O ports
add DM support for parallel I/O ports on QUICC Engine Block
Signed-off-by: Heiko Schocher <hs@denx.de mailto:hs@denx.de>
arch/powerpc/cpu/mpc83xx/cpu_init.c | 8 ++
arch/powerpc/cpu/mpc83xx/qe_io.c | 193
+++++++++++++++++++++++++++-
include/fsl_qe.h | 3 +
3 files changed, 201 insertions(+), 3 deletions(-)
diff --git a/arch/powerpc/cpu/mpc83xx/cpu_init.c
b/arch/powerpc/cpu/mpc83xx/cpu_init.c
index af8facad53..cfcc16607c 100644
--- a/arch/powerpc/cpu/mpc83xx/cpu_init.c
+++ b/arch/powerpc/cpu/mpc83xx/cpu_init.c
*//*
How about mpc85xx?
Yes, this is a Todo (one reason why this patch is RFC). I currently have reworked this for the mpc83xx based keymile (now abb) boards (which are locally running with complete DM/DTS support now).
You mean you have tested RFC patch on your side? That's great!
Yes, on the following boards:
"kmcoge5ne tuxx1 kmopti2 kmtegr1 tuge1 kmsupx5"
Not all in mainline (yet)
@Holger: should/could we mainline the missing ones?
This boards work with full DM/DTS support now, and have no compiletime warnign like:
===================== WARNING ====================== This board does not use CONFIG_DM_ETH (Driver Model
anymore...
Would you like to push the dts patch too?
Yes, but give me some time, until I can test the changed patchset again.
I have one mpx85xx based board I can try, but may others are working on this support, so I wanted to post my patches very early, so we can work together on it...
As my first question in patch comment is:
may we should move this part to drivers/pinctrl ?
Yes, it is more better to move to drivers/pinctrl.
Ok, I must look into it.
@@ -11,6 +11,9 @@
#ifdef CONFIG_USB_EHCI_FSL
#include <usb/ehci-ci.h>
#endif
+#ifdef CONFIG_QE
+#include <fsl_qe.h>
+#endif
*//*
If include this headfile, the function declaration for qe_init and qe_reset in this file should be removed as below:
extern void qe_init(uint qe_base);
extern void qe_reset(void);
Yes, of course! But this part of code I did not touch... may this needs a cleanup at all.
Just two lines extern declaration in this file.
I remove them in seperate patch.
#include "lblaw/lblaw.h"
#include "elbc/elbc.h"
@@ -27,6 +30,7 @@ extern void qe_config_iopin(u8 port, u8 pin, int dir,
extern void qe_init(uint qe_base); extern void qe_reset(void);
+/**
- par_io_of_config_node config
- @dev: pointer to pinctrl device
- @pio: ofnode of pinconfig property
- */
+static int par_io_of_config_node(struct udevice *dev, ofnode pio) {
+ struct qe_io_platdata *plat = dev->platdata;
+ volatile qepio83xx_t *par_io = plat->base;
+ const unsigned int *pio_map;
+ int pio_map_len;
+ pio_map = ofnode_get_property(pio, "pio-map", &pio_map_len);
+ if (pio_map == NULL)
+ return -ENOENT;
+ pio_map_len /= sizeof(unsigned int);
+ if ((pio_map_len % 6) != 0) {
+ printk(KERN_ERR "pio-map format wrong!\n");
+ return -EINVAL;
+ }
+ while (pio_map_len > 0) {
+ /*
+ * column pio_map[5] from linux (has_irq) not
+ * supported in u-boot yet.
+ */
*//*
remove or keep pio_map[5] in uboot dts, which is better?
I think we keep it as it is in linux dts. We simply ignore it (yet) in U-Boot.
That's OK.
+const struct pinctrl_ops par_io_pinctrl_ops = {
+ .set_state = par_io_pinctrl_set_state, };
+static const struct udevice_id par_io_pinctrl_match[] = {
+ { .compatible = "fsl,mpc8360-par_io"},
*//*
Why is*//*fsl,mpc8360-par_io, maybe fsl,qe-par-io are more better.*//*
Yes, more common, but in Linux is already:
https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kerne l.org%2Fpub%2Fscm%2Flinux%2Fkernel%2Fgit%2Ftorvalds%2Flinux.git%2Ftree %2Farch%2Fpowerpc%2Fboot%2Fdts%2Fkmeter1.dts%23n133&data=02% 7C01%7Cqiang.zhao%40nxp.com%7C783271f84fe043eaf21708d7e0f7df46%7 C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C637225227517025151 &sdata=5VZsJPI2QwZBnS5bgSDC1cTF1lKh%2BGQXHpP%2FF4jGzCU%3D& amp;reserved=0
or compatible = "fsl,mpc8323-qe-pario";
for board https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kerne l.org%2Fpub%2Fscm%2Flinux%2Fkernel%2Fgit%2Ftorvalds%2Flinux.git%2Ftree %2Farch%2Fpowerpc%2Fboot%2Fdts%2Fmpc832x_rdb.dts%23n163&data =02%7C01%7Cqiang.zhao%40nxp.com%7C783271f84fe043eaf21708d7e0f7df4 6%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C637225227517025 151&sdata=AwMW4NtqID9GwbM9RnZcqqDQVZ7RvuaRw5U6LFC3VWA% 3D&reserved=0
Unfortunately, it is more or less a mess in linux, as each board scans in board code for the node *name* not compatible entry ! See as an example:
https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgit.kerne l.org%2Fpub%2Fscm%2Flinux%2Fkernel%2Fgit%2Ftorvalds%2Flinux.git%2Ftree %2Farch%2Fpowerpc%2Fplatforms%2F83xx%2Fmpc832x_rdb.c%23n198& ;data=02%7C01%7Cqiang.zhao%40nxp.com%7C783271f84fe043eaf21708d7e0 f7df46%7C686ea1d3bc2b4c6fa92cd99c5c301635%7C0%7C0%7C6372252275 17025151&sdata=RG2pIeiGMFZwL16CLjRHWrvC12YBEDDi5H7kNoZXz0A %3D&reserved=0
So may we are free here, to add here a new compatible entry "fsl,qe-par-io" ? (But I want to be compatible with linux and accept the same compatible strings as linux has.)
OK!
Best Regards Qiang Zhao
bye, Heiko