malta: Support kernel 6.6 and fix kernel_bump.sh script #15215

guidosarducci · 2024-04-20T06:41:10Z

Adding kernel 6.6 support for malta has been tested and pending for more over a month in PR #14891 , but merging appears to be stalled. I've added this PR to tackle possibly outstanding issues and hopefully encourage some progress.

Additional changes from #14891 include:

Use scripts/kernel_bump.sh while adding malta 6.6 support to preserve history.
Fix scripts/kernel_bump.sh to generate commit text with correct word-wrapping.

@robimarko @PolynomialDivision @Ansuel @hauke Please feel free to pull anything needed from here into #14891, or simply merge this as is, or comment on further changes needed. I deferred my own malta/6.6 update expecting #14891 would be merged quickly, and would like to move things along since having 6.6 support helps with my upstream kernel work.

guidosarducci · 2024-04-21T00:19:58Z

Added commit to revert mips64 breakage due to c5dee97. See commit comments for more information.

This patch fixes the broken detect_memory_region() function on 6.6 kernel. Signed-off-by: Shiji Yang <yangshiji66@qq.com>

robimarko · 2024-04-21T09:48:33Z

@guidosarducci Has the revert been tested not to break other targets?

guidosarducci · 2024-04-21T15:59:23Z

@guidosarducci Has the revert been tested not to break other targets?

@robimarko I tested that both malta/mips32el and malta/mips64el work after the revert.

The original commit was part of #14774 (you asked about it too) and targets ramips but breaks all mips64 targets, and looks only compatible with 32-bit targets. I don't know ramips specifics so let's ask @DragonBluep for more details.

I would be fine with merging the malta update separately from dealing with this revert.

PolynomialDivision · 2024-04-22T09:11:27Z

The mentioned patch also targets ath79 see #14880.

DragonBluep · 2024-04-22T09:52:49Z

target/linux/generic/pending-6.6/350-mips-kernel-fix-detect_memory_region-function.patch

- void __init detect_memory_region(phys_addr_t start, phys_addr_t sz_min, phys_addr_t sz_max)
- {
-	void *dm = &detect_magic;
-+	void *dm = (void *)KSEG1ADDR(&detect_magic);


@namiltd Hi! Could you help test if changing this line to void *dm = &detect_magic; can work on your MT7620 device? I don't have Linux build environment recently.

Unfortunately, after this change, RAM size detection does not work correctly:

total used free shared buff/cache available Mem: 251372 21644 205884 232 23844 193116 Swap: 0 0 0

instead of:

total used free shared buff/cache available Mem: 56636 20804 11224 232 24608 11856 Swap: 0 0 0

This router only has 64MB RAM, but it incorrectly detects 256MB.

Why is the patch so complicated to begin with and lack a description of how/why it should work? Why does it write when the original code doesn't?

@namiltd Maybe try replacing the whole patch with something simpler:

--- a/arch/mips/kernel/setup.c +++ b/arch/mips/kernel/setup.c @@ -103,7 +103,7 @@ phys_addr_t size; for (size = sz_min; size < sz_max; size <<= 1) { - if (!memcmp(dm, dm + size, sizeof(detect_magic))) + if (*(long *)(dm) != *(volatile long __force *)(dm + size)) break; }

This at least compiles on 32-bit and 64-bit mips malta...

This patch (https://github.com/openwrt/openwrt/blob/main/target/linux/generic/pending-6.6/350-mips-kernel-fix-detect_memory_region-function.patch) was not caused by compilation errors, but because the router stopped during startup and displayed the following message:

[ 0.000000] Linux 6.6.20 (builder@buildhost) (mipsel-openwrt-linux-musl-gcc (OpenWrt GCC 12.3.0 r25430-eb44f39e39) 12.3.0, GNU ld (GNU Binutils) 2.40.0) Mar 6 11:40:57 2024 [ 0.000000] Board has DDR2 [ 0.000000] Analog PMU set to hw control [ 0.000000] Digital PMU set to hw control [ 0.000000] SoC Type: MediaTek MT7620A ver:2 eco:6 [ 0.000000] printk: bootconsole [early0] enabled [ 0.000000] CPU0 revision is: 00019650 (MIPS 24KEc) [ 0.000000] MIPS: machine is TP-Link Archer C5 v4 [ 0.000000] detected buffer overflow in memcmp

guidosarducci · 2024-04-22T22:27:24Z

The mentioned patch also targets ath79 see #14880.

@PolynomialDivision Yes, I'm aware. The upstream Linux patch history showed the reverted code was borrowed from existing ath79 implementation in 4d9f77d25268 ("MIPS: add detect_memory_region()") by @blogic . That doesn't really change anything however: the commit is still written for 32-bit only and thus wasn't well-tested, hasn't been reviewed upstream, and breaks mips64.

I assumed reverting this would be straightforward given the above but apparently not. Since I intended this PR to push along the stalled malta kernel 6.6 support rather than tangential development, maybe it's better to drop the revert and let any further testing/development continue in the original commit/PR by the folks involved.

namiltd · 2024-04-23T06:09:36Z

350-mips-kernel-fix-detect_memory_region-function.patch
@guidosarducci I think the easiest way was to modify this patch (https://github.com/openwrt/openwrt/blob/main/target/linux/generic/pending-6.6/350-mips-kernel-fix-detect_memory_region-function.patch) by detecting whether 32 or 64 bits and implementing them differently.

e.g.:

#ifdef CONFIG_CPU_MIPS64
static void *detect_magic __initdata = detect_memory_region;
#else
static u32 detect_magic __initdata;
#define MIPS_MEM_TEST_PATTERN		0xaa5555aa
#endif

#ifdef CONFIG_MIPS_AUTO_PFN_OFFSET
unsigned long ARCH_PFN_OFFSET;
EXPORT_SYMBOL(ARCH_PFN_OFFSET);
#endif

void __init detect_memory_region(phys_addr_t start, phys_addr_t sz_min, phys_addr_t sz_max)
{
	phys_addr_t size;
#ifdef CONFIG_CPU_MIPS64
	void *dm = &detect_magic;

	for (size = sz_min; size < sz_max; size <<= 1) {
		if (!memcmp(dm, dm + size, sizeof(detect_magic)))
			break;
	}
#else
	void* dm = (void *)KSEG1ADDR(&detect_magic); 

	for (size = sz_min; size < sz_max; size <<= 1) {
		__raw_writel(MIPS_MEM_TEST_PATTERN, dm);
		if (__raw_readl(dm) == __raw_readl(dm + size)) {
			__raw_writel(~MIPS_MEM_TEST_PATTERN, dm);
			if (__raw_readl(dm) == __raw_readl(dm + size))
				break;
		}
	}
#endif

	pr_debug("Memory: %lluMB of RAM detected at 0x%llx (min: %lluMB, max: %lluMB)\n",
		((unsigned long long) size) / SZ_1M,
		(unsigned long long) start,
		((unsigned long long) sz_min) / SZ_1M,
		((unsigned long long) sz_max) / SZ_1M);

	memblock_add(start, size);
}

New 350-mips-kernel-fix-detect_memory_region-function.patch

guidosarducci · 2024-04-24T00:17:27Z

@namiltd That seems very kludgy and it's difficult to see it being accepted upstream. The minimal patch I posted above has the original detect_memory_region() function behaviour but bypasses memcmp() by doing the equivalent of __raw_readl(). It also doesn't need any detection since sizeof(long) == sizeof(void *).

Could you try it on your device if you have a chance? i.e. just replace 350-mips-kernel-fix-detect_memory_region-function.patch with:

--- a/arch/mips/kernel/setup.c
+++ b/arch/mips/kernel/setup.c
@@ -103,7 +103,7 @@
        phys_addr_t size;

        for (size = sz_min; size < sz_max; size <<= 1) {
-               if (!memcmp(dm, dm + size, sizeof(detect_magic)))
+               if (*(long *)(dm) != *(volatile long __force *)(dm + size))
                        break;
        }

@DragonBluep Are you able to ping/bump your patch on the linux-mips mailing list to ask for advice?

guidosarducci · 2024-04-24T00:27:36Z

I've dropped the revert of c5dee97 as mentioned in #15215 (comment). That can be dealt with separately/in the the original PR.

@PolynomialDivision @robimarko @hauke It should be OK to merge this or incorporate needed changes into #14891, if you'd care to have a look. Thanks...

namiltd · 2024-04-24T06:31:04Z

@namiltd That seems very kludgy and it's difficult to see it being accepted upstream. The minimal patch I posted above has the original detect_memory_region() function behaviour but bypasses memcmp() by doing the equivalent of __raw_readl(). It also doesn't need any detection since sizeof(long) == sizeof(void *).

Could you try it on your device if you have a chance? i.e. just replace 350-mips-kernel-fix-detect_memory_region-function.patch with:
--- a/arch/mips/kernel/setup.c
+++ b/arch/mips/kernel/setup.c
@@ -103,7 +103,7 @@
        phys_addr_t size;

        for (size = sz_min; size < sz_max; size <<= 1) {
-               if (!memcmp(dm, dm + size, sizeof(detect_magic)))
+               if (*(long *)(dm) != *(volatile long __force *)(dm + size))
                        break;
        }

@guidosarducci I don't have physical access to the mt7620 router for the next two weeks, so I won't test it.

I took the risk of updating remotely and unfortunately the router did not get up after the update. That's all I can say without having physical access to it. On site, the network was connected to another router and I will restore it to the previous version at a later date.

guidosarducci · 2024-04-26T00:41:11Z

@guidosarducci I don't have physical access to the mt7620 router for the next two weeks, so I won't test it.

I took the risk of updating remotely and unfortunately the router did not get up after the update. That's all I can say without having physical access to it. On site, the network was connected to another router and I will restore it to the previous version at a later date.

@namiltd Would you clarify if it was a "regular" update that failed to boot, or one with the last patch I posted? Timing of your post and then edit was a little confusing...

DragonBluep · 2024-04-26T13:10:55Z

@namiltd @guidosarducci A workaround for mips64 ARCH:
0001-generic-MIPS64-fix-detect_memory_region-compilation-.patch

Text of a commit message body should wrap at 75 characters. Manual commits are expected to do so, but automated commits *must* do so to avoid adding repeated ugly commits. Signed-off-by: Tony Ambardar <itugrok@yahoo.com>

This is an automatically generated commit. When doing `git bisect`, consider `git bisect --skip`. Signed-off-by: Tony Ambardar <itugrok@yahoo.com>

This is an automatically generated commit which aids following Kernel patch history, as git will see the move and copy as a rename thus defeating the purpose. For the original discussion see: https://lists.openwrt.org/pipermail/openwrt-devel/2023-October/041673.html Signed-off-by: Tony Ambardar <itugrok@yahoo.com>

This refreshes the configuration on top of kernel 6.6 and activates it as test kernel configuration. Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de> Signed-off-by: Tony Ambardar <itugrok@yahoo.com>

hauke · 2024-04-28T22:27:08Z

Thank you for taking car of this PR and sorry for the delay.
@DragonBluep Could you please create a separate PR with the fix.

github-actions bot added build/scripts/tools pull request/issues for build, scripts and tools related changes kernel pull request/issue with Linux kernel related changes target/malta pull request/issue for malta target labels Apr 20, 2024

namiltd mentioned this pull request Apr 20, 2024

kernel: status of 6.6 version #15192

Open

42 tasks

guidosarducci force-pushed the master-add-malta-6.6 branch from ec5f4c6 to 0f650d9 Compare April 21, 2024 00:15

guidosarducci changed the title ~~malta: Add kernel 6.6 support and fix kernel_bump.sh script~~ malta: Add kernel 6.6 support, fix kernel_bump.sh script, revert mips64 breakage Apr 21, 2024

guidosarducci changed the title ~~malta: Add kernel 6.6 support, fix kernel_bump.sh script, revert mips64 breakage~~ malta: Support kernel 6.6, fix kernel_bump.sh script, revert mips64 breakage Apr 21, 2024

guidosarducci referenced this pull request Apr 21, 2024

generic: fix detect_memory_region() function

c5dee97

This patch fixes the broken detect_memory_region() function on 6.6 kernel. Signed-off-by: Shiji Yang <yangshiji66@qq.com>

DragonBluep reviewed Apr 22, 2024

View reviewed changes

guidosarducci force-pushed the master-add-malta-6.6 branch from 0f650d9 to f919520 Compare April 24, 2024 00:19

guidosarducci changed the title ~~malta: Support kernel 6.6, fix kernel_bump.sh script, revert mips64 breakage~~ malta: Support kernel 6.6 and fix kernel_bump.sh script Apr 24, 2024

namiltd mentioned this pull request Apr 24, 2024

ramips: add Wavlink WL-WN572HG3 router #15179

Open

guidosarducci mentioned this pull request Apr 26, 2024

multiple related updates: gnulib, coreutils, gettext-full, elfutils #14802

Merged

guidosarducci and others added 4 commits April 29, 2024 00:24

scripts/kernel_bump: Fix commit text formatting

66252aa

Text of a commit message body should wrap at 75 characters. Manual commits are expected to do so, but automated commits *must* do so to avoid adding repeated ugly commits. Signed-off-by: Tony Ambardar <itugrok@yahoo.com>

kernel/malta: Create kernel files for v6.6 (from v6.1)

3d0bf90

This is an automatically generated commit. When doing `git bisect`, consider `git bisect --skip`. Signed-off-by: Tony Ambardar <itugrok@yahoo.com>

malta: refresh kernel 6.6 configuration

a638d01

This refreshes the configuration on top of kernel 6.6 and activates it as test kernel configuration. Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de> Signed-off-by: Tony Ambardar <itugrok@yahoo.com>

hauke force-pushed the master-add-malta-6.6 branch from f919520 to a638d01 Compare April 28, 2024 22:25

openwrt-bot merged commit a638d01 into openwrt:main Apr 28, 2024
3 checks passed

hauke mentioned this pull request Apr 28, 2024

malta: Add kernel 6.6 support #14891

Closed

guidosarducci deleted the master-add-malta-6.6 branch April 29, 2024 00:01

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

malta: Support kernel 6.6 and fix kernel_bump.sh script #15215

malta: Support kernel 6.6 and fix kernel_bump.sh script #15215

guidosarducci commented Apr 20, 2024 •

edited

guidosarducci commented Apr 21, 2024

robimarko commented Apr 21, 2024

guidosarducci commented Apr 21, 2024 •

edited

PolynomialDivision commented Apr 22, 2024

DragonBluep Apr 22, 2024

namiltd Apr 22, 2024 •

edited

guidosarducci Apr 22, 2024 •

edited

namiltd Apr 23, 2024 •

edited

guidosarducci commented Apr 22, 2024

namiltd commented Apr 23, 2024 •

edited

guidosarducci commented Apr 24, 2024

guidosarducci commented Apr 24, 2024

namiltd commented Apr 24, 2024 •

edited

guidosarducci commented Apr 26, 2024

DragonBluep commented Apr 26, 2024

hauke commented Apr 28, 2024

malta: Support kernel 6.6 and fix kernel_bump.sh script #15215

malta: Support kernel 6.6 and fix kernel_bump.sh script #15215

Conversation

guidosarducci commented Apr 20, 2024 • edited

guidosarducci commented Apr 21, 2024

robimarko commented Apr 21, 2024

guidosarducci commented Apr 21, 2024 • edited

PolynomialDivision commented Apr 22, 2024

DragonBluep Apr 22, 2024

Choose a reason for hiding this comment

namiltd Apr 22, 2024 • edited

Choose a reason for hiding this comment

guidosarducci Apr 22, 2024 • edited

Choose a reason for hiding this comment

namiltd Apr 23, 2024 • edited

Choose a reason for hiding this comment

guidosarducci commented Apr 22, 2024

namiltd commented Apr 23, 2024 • edited

guidosarducci commented Apr 24, 2024

guidosarducci commented Apr 24, 2024

namiltd commented Apr 24, 2024 • edited

guidosarducci commented Apr 26, 2024

DragonBluep commented Apr 26, 2024

hauke commented Apr 28, 2024

guidosarducci commented Apr 20, 2024 •

edited

guidosarducci commented Apr 21, 2024 •

edited

namiltd Apr 22, 2024 •

edited

guidosarducci Apr 22, 2024 •

edited

namiltd Apr 23, 2024 •

edited

namiltd commented Apr 23, 2024 •

edited

namiltd commented Apr 24, 2024 •

edited