rp2470 boot crash in post-5.8 git

While 5.8.0 worked fine on my HP RP2470 (Gentoo with 10.2.0 kgcc), yesterdays dddcbc139e96 and todays a1d21081a60d crash on boot. I have started bisecting it.

Last output lines from 5.8.0-13161-gdddcbc139e96:

[    1.150985] 1. Crescendo 750 W2 at 0xfffffffffffa0000 [160] { 0, 0x0, 0x5e3, 0x00004 }
[    1.243570] 2. Memory at 0xfffffffffed08000 [8] { 1, 0x0, 0x09b, 0x00009 }
[    1.243792] 3. Astro BC Runway Port at 0xfffffffffed00000 [0] { 12, 0x0, 0x582, 0x0000b }
[    1.382893] 4. Elroy PCI Bridge at 0xfffffffffed30000 [0/0] { 13, 0x0, 0x782, 0x0000a }
[    1.522243] 5. Elroy PCI Bridge at 0xfffffffffed34000 [0/2] { 13, 0x0, 0x782, 0x0000a }
[    1.660767] 6. Elroy PCI Bridge at 0xfffffffffed38000 [0/4] { 13, 0x0, 0x782, 0x0000a }
[    1.799247] 7. Elroy PCI Bridge at 0xfffffffffed3c000 [0/6] { 13, 0x0, 0x782, 0x0000a }
[    1.799500]
********** VIRTUAL FRONT PANEL **********
System Boot detected
*****************************************
LEDs:  RUN      ATTENTION     FAULT     REMOTE     POWER
       ON       FLASH         OFF       ON         ON

LED State: There was a system interruption that did not take the system down. Check Chassis and Console Logs for error messages.

processor system initialization 1C00

*****************************************

************ EARLY BOOT VFP *************
End of early boot detected
*****************************************
h support not available.

[    2.639001] HugeTLB registered 2.00 MiB page size, pre-allocated 0 pages

********** VIRTUAL FRONT PANEL **********
System Boot detected
*****************************************

LEDs:  RUN      ATTENTION     FAULT     REMOTE     POWER
       ON       FLASH         FLASH     ON         ON

LED State: System Running. Unexpected Reboot. Non-critical Error Detected. Check Chassis and Console Logs for error messages.

processor system panic 1B00

---------------------------------------------------------------------------------------------------
Last output lines from 5.8.0-13249-ga1d21081a60d:

[    1.151696] 1. Crescendo 750 W2 at 0xfffffffffffa0000 [160] { 0, 0x0, 0x5e3, 0x00004 }
[    1.242819] 2. Memory at 0xfffffffffed08000 [8] { 1, 0x0, 0x09b, 0x00009 }
[    1.243041] 3. Astro BC Runway Port at 0xfffffffffed00000 [0] { 12, 0x0, 0x582, 0x0000b }
[    1.383414] 4. Elroy PCI Bridge at 0xfffffffffed30000 [0/0] { 13, 0x0, 0x782, 0x0000a }
[    1.523072] 5. Elroy PCI Bridge at 0xfffffffffed34000 [0/2] { 13, 0x0, 0x782, 0x0000a }
[    1.662053] 6. Elroy PCI Bridge at 0xfffffffffed38000 [0/4] { 13, 0x0, 0x782, 0x0000a }
[    1.800405] 7. Elroy PCI Bridge at 0xfffffffffed3c000 [0/6] { 13, 0x0, 0x782, 0x0000a }
[    1.800658]
********** VIRTUAL FRONT PANEL **********
System Boot detected
*****************************************
LEDs:  RUN      ATTENTION     FAULT     REMOTE     POWER
       ON       FLASH         OFF       ON         ON

LED State: There was a system interruption that did not take the system down. Check Chassis and Console Logs for error messages.

processor system initialization 1C00

*****************************************

************ EARLY BOOT VFP *************
End of early boot detected
*****************************************

********** VIRTUAL FRONT PANEL **********
System Boot detected
*****************************************

LEDs:  RUN      ATTENTION     FAULT     REMOTE     POWER
       ON       FLASH         FLASH     ON         ON

LED State: System Running. Unexpected Reboot. Non-critical Error Detected. Check Chassis and Console Logs for error messages.

processor system panic 1B00

*****************************************

************ EARLY BOOT VFP *************
End of early boot detected
*****************************************

********** VIRTUAL FRONT PANEL **********
System Boot detected
*****************************************

LEDs:  RUN      ATTENTION     FAULT     REMOTE     POWER
       ON       FLASH         FLASH     ON         ON

LED State: System Running. Unexpected Reboot. Non-critical Error Detected. Check Chassis and Console Logs for error messages.

processor system panic 1B00

Meelis Roos <[email protected]>

And finished now. I included all CC-s and SOB-s from the commit to this mail - sorry if too broad.

1355c31eeb7ea61a7f2f2937d17cd4e343a6b5af is the first bad commit commit 1355c31eeb7ea61a7f2f2937d17cd4e343a6b5af

Author: Mike Rapoport <[email protected]>
Date:   Thu Aug 6 23:22:39 2020 -0700

asm-generic: pgalloc: provide generic pmd_alloc_one() and pmd_free_one()

For most architectures that support >2 levels of page tables, pmd_alloc_one() is a wrapper for __get_free_pages(), sometimes with __GFP_ZERO and sometimes followed by memset(0) instead.

More elaborate versions on arm64 and x86 account memory for the user page tables and call to pgtable_pmd_page_ctor() as the part of PMD page initialization.

Move the arm64 version to include/asm-generic/pgalloc.h and use the generic version on several architectures.

The pgtable_pmd_page_ctor() is a NOP when ARCH_ENABLE_SPLIT_PMD_PTLOCK is not enabled, so there is no functional change for most architectures except of the addition of __GFP_ACCOUNT for allocation of user page tables.

The pmd_free() is a wrapper for free_page() in all the cases, so no functional change here.

 arch/alpha/include/asm/pgalloc.h     | 15 +------------
 arch/arm/include/asm/pgalloc.h       | 11 ---------
 arch/arm64/include/asm/pgalloc.h     | 27 +---------------------
 arch/ia64/include/asm/pgalloc.h      | 10 ---------
 arch/mips/include/asm/pgalloc.h      |  8 ++-----
 arch/parisc/include/asm/pgalloc.h    | 11 ++-------
 arch/riscv/include/asm/pgalloc.h     | 13 +----------
 arch/sh/include/asm/pgalloc.h        |  3 +++
 arch/um/include/asm/pgalloc.h        |  8 +------
 arch/um/include/asm/pgtable-3level.h |  3 ---
 arch/um/kernel/mem.c                 | 12 ----------
 arch/x86/include/asm/pgalloc.h       | 26 +---------------------
 include/asm-generic/pgalloc.h        | 43 ++++++++++++++++++++++++++++++++++++
 13 files changed, 55 insertions(+), 135 deletions(-)

Meelis Roos <[email protected]>

yes, I see the problem.

arch/parisc/include/asm/pgtable.h:#define PMD_ORDER 1 /* Number of pages per pmd */

-static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long address) -{
- pmd_t *pmd = (pmd_t *)__get_free_pages(GFP_KERNEL, PMD_ORDER);

+ page = alloc_pages(gfp, 0);

So I think the quick fix is to restore the parisc pmd_alloc_one and define __HAVE_ARCH_PMD_ALLOC_ONE. I wasn't following this series too closely, so I shan't propose a patch, but leave it up to Mike. I'd probably miss some subtlety.

Hi,

You are right, I've missed that parisc has PMD_ORDER != 0 and the simplest fix would be to revert changes to parisc pmd_alloc_one().

Meelis, can you please try the below patch:

index cc7ecc2ef55d..a6482b2ce0ea 100644
@@ -10,6 +10,7 @@
 
 #include <asm/cache.h>
 
+#define __HAVE_ARCH_PMD_ALLOC_ONE
 #define __HAVE_ARCH_PMD_FREE
 #define __HAVE_ARCH_PGD_FREE
 #include <asm-generic/pgalloc.h>

@@ -67,6 +68,11 @@ static inline void pud_populate(struct mm_struct *mm, pud_t *pud, pmd_t *pmd)
 			(__u32)(__pa((unsigned long)pmd) >> PxD_VALUE_SHIFT)));
 }
 
+static inline pmd_t *pmd_alloc_one(struct mm_struct *mm, unsigned long address)
+{
+	return (pmd_t *)__get_free_pages(GFP_PGTABLE_KERNEL, PMD_ORDER);
+}
+
 static inline void pmd_free(struct mm_struct *mm, pmd_t *pmd)
 {
 	if (pmd_flag(*pmd) & PxD_FLAG_ATTACHED) {

Sincerely yours,
Mike.

[...]

It works on top of current git 5.8.0-13249-ga1d21081a60d - thank you!