Enable I2C#6 instead of I2C#8 #12

staroselskii · 2018-08-30T14:13:00Z

Initial I2C#6 pins' configuration came from the firmware. Which quite likely has been used in the phones, where I2C#8, that is not part of Atom peripheral, is in use. Thus we need to override the leftover.

In order to achieve that I needed to port some of your Linux kernel code to U-Boot:

scu_ipc_raw_command() that takes care of the complex command sent to MCU to let go off of I2C#6 pins
add an adapted version of pinctrl-merrifield.c in pinmux.c that takes care of the above mentioned pins
add an ACPI binding to make it usable by others

andy-shev · 2018-08-30T14:49:58Z

Almost there, I would leave few comments, feel free to address them and resend this PR.

P.S. Looks pretty much cool! Thanks for your work!

andy-shev · 2018-08-30T14:50:54Z

arch/x86/cpu/tangier/pinmux.c

+	}
+}
+
+int mrfld_pinconfig_protected(unsigned int pin, u32 mask, u32 bits)


Since this is an API it has to get a pin control device first, something like

ret = syscon_get_by_driver_data(X86_SYSCON_SCU, &dev);

The idea behind is to get lazy probe (that's called "Driver Model" in U-Boot).

@andy-shev
Thanks. Makes sense.

But it seems like probe is already automagically called just by the sheer addition of the probe callback, doesn't it? I'm pretty sure I'm missing something, of course.

If you call function before device is probed, it wouldn't work and crash everything. ->probe() will be called when we try to get pointer to a device.

andy-shev · 2018-08-30T14:52:14Z

arch/x86/cpu/tangier/pinmux.c

+
+static int tangier_pinctrl_probe(struct udevice *dev)
+{
+    void *base_addr = syscon_get_first_range(X86_SYSCON_PINCONF);


Mind indentation.

andy-shev · 2018-08-30T14:54:32Z

arch/x86/cpu/tangier/pinmux.c

+
+int mrfld_pinconfig_protected(unsigned int pin, u32 mask, u32 bits)
+{
+	struct mrfld_family *mf = &i2c_family;


You need to keep your families somewhere else and here convert pin to family followed by family to bufcfg address.

Okay. Something like mrfld_get_family from pinctrl-merrifield.c?

andy-shev · 2018-08-30T14:55:14Z

arch/x86/cpu/tangier/pinmux.c

+}
+
+static const struct udevice_id tangier_pinctrl_match[] = {
+	{ .compatible = "intel,pinctrl-tangier",


One line?
I mean compatible string with second parameter can be on one line. can they?

andy-shev · 2018-08-30T14:55:59Z

arch/x86/cpu/tangier/pinmux.c

+
+	ret = scu_ipc_raw_command(cmd, 0, &v, 4, NULL, 0, (u32) bufcfg, 0);
+	if (ret)
+		printf("Failed to set bufcfg via SCU IPC for pin %u (%d)\n", pin, ret);


Hmm... I don't remember if err() stops an execution, please check and use it instead.

andy-shev · 2018-08-30T14:57:20Z

arch/x86/cpu/tangier/pinmux.c

+
+	v = (value & ~mask) | (bits & mask);
+
+	debug("scu: cmd: %d v: 0x%x p: 0x%x bits: %d, mask: %d bufcfg: 0x%p\n",


I doubt it's useful in production. (Note ahead, I'm wondering if upstream will ask to use regmap for this driver, which, I guess, has a facility to dump registers)

Not useful but this was extremely useful for me to get the parameters right. I'll get rid of it as soon as we get everything else.

Fine to me as well, but should be removed before sending upstream.

andy-shev · 2018-08-30T14:59:24Z

arch/x86/cpu/tangier/pinmux.c

@@ -0,0 +1,100 @@
+// SPDX-License-Identifier: GPL-2.0


Yep, like I said in the bug report, check pinctrl-sandbox.c for a hint how this should look like. (Yes, I understand that we are creating a limited driver, it needs to be taken into account)

Okay. So we're making it the UCLASS_PINCTRL, then? With some of pinctrl_ops implemented? And then we'll configure the pins in edison.c using the pinctrl API? Is this the plan? If so, I'll need some time adapting this code to the pinctrl API.

I'm pretty sure that this what upstream ask us during review. However, we can try first with reduced solution. I guess, we may send it upstream and see what they will tell. In any case it would be nice to have your stuff in my tree sooner.

This interface will be used to configure properly some pins on Merrifield that are shared with SCU. scu_ipc_raw_command() writes SPTR and DPTR registers before sending a command to SCU. This code has been ported from Linux work done by Andy Shevchenko. Signed-off-by: Georgii Staroselskii <georgii.staroselskii@emlid.com>

staroselskii · 2018-08-30T22:17:58Z

@andy-shev
Hello, Andy!

I addressed most of your concerns and made a force push. I also fixed most of the warnings checkpatch.pl gave along the way.

andy-shev · 2018-08-31T06:02:15Z

arch/x86/cpu/tangier/pinmux.c

+	unsigned int bufno;
+
+	family =  mrfld_get_family(pinctrl, pin);
+


Redundant blank line

andy-shev · 2018-08-31T06:04:50Z

arch/x86/cpu/tangier/pinmux.c

+};
+
+static const struct mrfld_family *mrfld_get_family(struct mrfld_pinctrl *mp,
+						   unsigned int pin)


I dunno if it's GH web or in your sources you have this indentation issue (on the second line u should be in the same column as s on the line above)

Another possibility is to split like

static const struct mrfld_family * mrfld_get_fmaily(...)

andy-shev · 2018-08-31T06:05:07Z

arch/x86/cpu/tangier/pinmux.c

+}
+
+static void __iomem *mrfld_get_bufcfg(struct mrfld_pinctrl *pinctrl,
+			unsigned int pin)


Same about indentation

andy-shev · 2018-08-31T06:05:26Z

arch/x86/cpu/tangier/pinmux.c

+}
+
+static void mrfld_setup_families(void *base_addr, struct mrfld_family *families,
+			unsigned int nfam)


Same with indentation.

andy-shev · 2018-08-31T06:06:54Z

arch/x86/cpu/tangier/pinmux.c

+	v = (value & ~mask) | (bits & mask);
+
+	debug("scu: v: 0x%x p: 0x%x bits: %d, mask: %d bufcfg: 0x%p\n",
+			v, (u32)bufcfg, bits, mask, bufcfg);


Indentation.

andy-shev · 2018-08-31T06:07:01Z

arch/x86/cpu/tangier/pinmux.c

+			v, (u32)bufcfg, bits, mask, bufcfg);
+
+	ret = scu_ipc_raw_command(IPCMSG_INDIRECT_WRITE, 0, &v, 4,
+			NULL, 0, (u32)bufcfg, 0);


andy-shev · 2018-08-31T06:07:15Z

arch/x86/cpu/tangier/pinmux.c

+	struct mrfld_pinctrl *pinctrl = dev_get_priv(dev);
+
+	mrfld_setup_families(base_addr, mrfld_families,
+			ARRAY_SIZE(mrfld_families));


andy-shev · 2018-08-31T06:10:06Z

Please, address some minor issues and I'm ready to take it.

This API is going to be used to configure some pins that are protected for simple modification. It's not a comprehensive pinctrl driver but can be turned into one when we need this in the future. Now it is planned to be used only in one place. So that's why I decided not to polute the codebase with a full-blown pinctrl-merrifield nobody will use. The code has been ported from Linux work done by Andy Shevchenko in pinctrl-merrfifield.c Signed-off-by: Georgii Staroselskii <georgii.staroselskii@emlid.com>

Initial configuration came from the firmware. Which quite likely has been used in the phones, where I2C #8, that is not part of Atom peripheral, is in use. Thus we need to override the leftover. In order to achieve this goal we make use of the new code in pinmux.c and set the appropriate pin function. From now on we can use I2C#6 in Linux kernel. Signed-off-by: Georgii Staroselskii <georgii.staroselskii@emlid.com>

Now that we have I2C#6 working, it's time to add a ACPI binding. Signed-off-by: Georgii Staroselskii <georgii.staroselskii@emlid.com>

staroselskii · 2018-08-31T10:42:50Z

@andy-shev

Hi, I think I got it this time.

I set ts=8 sw=8 and see that I have scu_ipc_raw_command() properly aligned in Vim but GH shows them a little unaligned. Is this an issue?

Compiling U-Boot with ubsan/asan libraries and running it in sandbox may lead to below backtrace: => avb init 0 => avb verify ## Android Verified Boot 2.0 version 1.1.0 read_is_device_unlocked not supported yet common/avb_verify.c:407:31: runtime error: division by zero AddressSanitizer:DEADLYSIGNAL Reviewed-by: Igor Opaniuk <igor.opaniuk@linaro.org> ================================================================= ==9388==ERROR: AddressSanitizer: FPE on unknown address 0x0000004b467f \ (pc 0x0000004b467f bp 0x000000000000 sp 0x7ffd899fe150 T0) #0 0x4b467e in mmc_byte_io common/avb_verify.c:407 #1 0x4b4c47 in mmc_byte_io common/avb_verify.c:532 #2 0x4b4c47 in read_from_partition common/avb_verify.c:533 #3 0x69dc0d in load_and_verify_vbmeta lib/libavb/avb_slot_verify.c:560 andy-shev#4 0x6a1ee6 in avb_slot_verify lib/libavb/avb_slot_verify.c:1139 #5 0x45dabd in do_avb_verify_part cmd/avb.c:245 #6 0x4af77c in cmd_call common/command.c:499 andy-shev#7 0x4af77c in cmd_process common/command.c:538 #8 0x46bafc in run_pipe_real common/cli_hush.c:1677 #9 0x46bafc in run_list_real common/cli_hush.c:1875 #10 0x46c780 in run_list common/cli_hush.c:2024 #11 0x46c780 in parse_stream_outer common/cli_hush.c:3216 andy-shev#12 0x46d34b in parse_file_outer common/cli_hush.c:3299 andy-shev#13 0x4ad609 in cli_loop common/cli.c:217 andy-shev#14 0x4625ae in main_loop common/main.c:65 u-boot#15 0x46f2d1 in run_main_loop common/board_r.c:648 u-boot#16 0x640253 in initcall_run_list lib/initcall.c:30 u-boot#17 0x46f9d0 in board_init_r common/board_r.c:879 u-boot#18 0x40539b in main arch/sandbox/cpu/start.c:321 u-boot#19 0x7fa94925f82f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2082f) u-boot#20 0x408908 in _start (/srv/R/u-boot-master/u-boot+0x408908) AddressSanitizer can not provide additional info. SUMMARY: AddressSanitizer: FPE common/avb_verify.c:407 in mmc_byte_io ==9388==ABORTING Signed-off-by: Eugeniu Rosca <erosca@de.adit-jv.com>

On aarch64 running with dcache off, will result in an unaligned access exception: => dcache off => hash sha1 $kernel_addr_r 100 "Synchronous Abort" handler, esr 0x96000061 elr: 00000000960317d8 lr : 00000000960316a4 (reloc) elr: 00000000fbd787d8 lr : 00000000fbd786a4 [..] The compiler emits a "stur x1, [x0, #12]". x1 is might just be 32 bit aligned pointer. Remove the unused u64 element from the union to drop the minimal alignment to 32 bit. Also remove the union, because it is no more needed. Signed-off-by: Michael Walle <michael@walle.cc> Reviewed-by: Priyanka Jain <priyanka.jain@nxp.com>

andy-shev requested changes Aug 30, 2018

View reviewed changes

staroselskii force-pushed the pr-add-edison-i2c-6-pinmux-v3 branch 2 times, most recently from 4ee1a06 to d9f450c Compare August 30, 2018 21:54

staroselskii force-pushed the pr-add-edison-i2c-6-pinmux-v3 branch from d9f450c to 4936fa1 Compare August 30, 2018 22:16

staroselskii mentioned this pull request Aug 30, 2018

Current master doesn't boot? edison-fw/meta-intel-edison#37

Closed

andy-shev requested changes Aug 31, 2018

View reviewed changes

staroselskii added 3 commits August 31, 2018 13:39

x86: tangier: acpi: add I2C6 node

869c471

Now that we have I2C#6 working, it's time to add a ACPI binding. Signed-off-by: Georgii Staroselskii <georgii.staroselskii@emlid.com>

staroselskii force-pushed the pr-add-edison-i2c-6-pinmux-v3 branch from 4936fa1 to 869c471 Compare August 31, 2018 10:39

andy-shev approved these changes Aug 31, 2018

View reviewed changes

andy-shev merged commit 749d17b into andy-shev:edison-acpi Aug 31, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Enable I2C#6 instead of I2C#8 #12

Enable I2C#6 instead of I2C#8 #12

staroselskii commented Aug 30, 2018 •

edited

Loading

andy-shev commented Aug 30, 2018

andy-shev Aug 30, 2018

staroselskii Aug 30, 2018

andy-shev Aug 30, 2018

andy-shev Aug 30, 2018

andy-shev Aug 30, 2018

staroselskii Aug 30, 2018

andy-shev Aug 30, 2018

andy-shev Aug 30, 2018 •

edited

Loading

andy-shev Aug 30, 2018

andy-shev Aug 30, 2018

staroselskii Aug 30, 2018

andy-shev Aug 30, 2018

andy-shev Aug 30, 2018

staroselskii Aug 30, 2018 •

edited

Loading

andy-shev Aug 30, 2018

staroselskii commented Aug 30, 2018

andy-shev Aug 31, 2018

andy-shev Aug 31, 2018 •

edited

Loading

andy-shev Aug 31, 2018

andy-shev Aug 31, 2018

andy-shev Aug 31, 2018

andy-shev Aug 31, 2018

andy-shev Aug 31, 2018

andy-shev commented Aug 31, 2018

staroselskii commented Aug 31, 2018


		v = (value & ~mask) \| (bits & mask);

		debug("scu: cmd: %d v: 0x%x p: 0x%x bits: %d, mask: %d bufcfg: 0x%p\n",

Enable I2C#6 instead of I2C#8 #12

Enable I2C#6 instead of I2C#8 #12

Conversation

staroselskii commented Aug 30, 2018 • edited Loading

andy-shev commented Aug 30, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andy-shev Aug 30, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

staroselskii Aug 30, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

staroselskii commented Aug 30, 2018

Choose a reason for hiding this comment

andy-shev Aug 31, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

andy-shev commented Aug 31, 2018

staroselskii commented Aug 31, 2018

staroselskii commented Aug 30, 2018 •

edited

Loading

andy-shev Aug 30, 2018 •

edited

Loading

staroselskii Aug 30, 2018 •

edited

Loading

andy-shev Aug 31, 2018 •

edited

Loading