"mm: consolidate pte_index() and pte_offset_*() definitions" broke ia64
From: John Paul Adrian Glaubitz
Date: Tue Aug 11 2020 - 12:35:22 EST
Hi Mike!
I just bisected a kernel issue on ia64 which leads to the kernel hanging very early
when booting on an HP RX2600 server (also verified to hang on other ia64 machines):
Loading Linux 5.8.0-12299-g00e4db51259a ...
Loading initial ramdisk ...
[ 0.000000] Linux version 5.8.0-12299-g00e4db51259a (root@glendronach) (gcc (Debian 10.2.0-3) 10.2.0, GNU ld (GNU Binutils for Debian) 2.35) #5 SMP Tue Aug 11 15:33:11 CEST 2020
[ 0.000000] efi: EFI v2.00 by HP
[ 0.000000] efi: SALsystab=0x3ee7a000 ACPI 2.0=0x3fde4000 ESI=0x3ee7b000 SMBIOS=0x3ee7c000 HCDP=0x3fde2000
[ 0.000000] PCDP: v3 at 0x3fde2000
[ 0.000000] earlycon: uart8250 at MMIO 0x0000000088033000 (options '115200n8')
[ 0.000000] printk: bootconsole [uart8250] enabled
[ 0.000000] ACPI: Early table checksum verification disabled
[ 0.000000] ACPI: RSDP 0x000000003FDE4000 000028 (v02 HP )
[ 0.000000] ACPI: XSDT 0x000000003FDE402C 0000A4 (v01 HP rx2660 00000000 HP 00000000)
[ 0.000000] ACPI: FACP 0x000000003FDF6A08 0000F4 (v03 HP rx2660 00000000 HP 00000000)
[ 0.000000] ACPI: DSDT 0x000000003FDE41C8 00E566 (v01 HP rx2660 00000007 INTL 20050309)
[ 0.000000] ACPI: FACS 0x000000003FDF6B00 000040
[ 0.000000] ACPI: SPCR 0x000000003FDF6B40 000050 (v01 HP 00000000 HP 00000000)
[ 0.000000] ACPI: DBGP 0x000000003FDF6B90 000034 (v01 HP rx2660 00000000 HP 00000000)
[ 0.000000] ACPI: APIC 0x000000003FDF6FB0 0000C8 (v01 HP rx2660 00000000 HP 00000000)
[ 0.000000] ACPI: SPMI 0x000000003FDF6BC8 000050 (v04 HP rx2660 00000000 HP 00000000)
[ 0.000000] ACPI: CPEP 0x000000003FDF6E80 000034 (v01 HP rx2660 00000000 HP 00000000)
[ 0.000000] ACPI: SSDT 0x000000003FDF2738 0004B3 (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF2BF8 000456 (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF3058 000EB8 (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF3F18 000EB8 (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF4DD8 000866 (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF5648 000EB8 (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF6508 000138 (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF6648 00013C (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF6788 00013C (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: SSDT 0x000000003FDF68C8 00013C (v01 HP rx2660 00000006 INTL 20050309)
[ 0.000000] ACPI: Local APIC address (____ptrval____)
[ 0.000000] 4 CPUs available, 4 CPUs total
[ 0.000000] SMP: Allowing 4 CPUs, 0 hotplug CPUs
[ 0.000000] Initial ramdisk at: 0xe00000002e368000 (9818100 bytes)
[ 0.000000] SAL 3.20: HP version 4.4
[ 0.000000] SAL Platform features:
[ 0.000000] None
[ 0.000000] SAL: AP wakeup using external interrupt vector 0xff
[ 0.000000] MCA related initialization done
[ 0.000000] Virtual mem_map starts at 0x(____ptrval____)
[ 0.000000] Zone ranges:
[ 0.000000] DMA32 [mem 0x0000000001000000-0x00000000ffffffff]
[ 0.000000] Normal [mem 0x0000000100000000-0x000001007fffffff]
[ 0.000000] Movable zone start for each node
[ 0.000000] Early memory node ranges
[ 0.000000] node 0: [mem 0x0000000001000000-0x000000003e67ffff]
[ 0.000000] node 0: [mem 0x000000003eaec000-0x000000003ee77fff]
[ 0.000000] node 0: [mem 0x000000003fc00000-0x000000003fd77fff]
[ 0.000000] node 0: [mem 0x000000003fddc000-0x000000003fddffff]
[ 0.000000] node 0: [mem 0x0000010040000000-0x000001007f1fbfff]
[ 0.000000] node 0: [mem 0x000001007f200000-0x000001007fffffff]
[ 0.000000] Initmem setup node 0 [mem 0x0000000001000000-0x000001007fffffff]
Bisecting the problem lead to your change as mentioned in the topic:
974b9b2c68f3d35a65e80af9657fe378d2439b60 is the first bad commit
commit 974b9b2c68f3d35a65e80af9657fe378d2439b60
Author: Mike Rapoport <rppt@xxxxxxxxxxxxx>
Date: Mon Jun 8 21:33:10 2020 -0700
mm: consolidate pte_index() and pte_offset_*() definitions
All architectures define pte_index() as
(address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1)
and all architectures define pte_offset_kernel() as an entry in the array
of PTEs indexed by the pte_index().
For the most architectures the pte_offset_kernel() implementation relies
on the availability of pmd_page_vaddr() that converts a PMD entry value to
the virtual address of the page containing PTEs array.
Let's move x86 definitions of the PTE accessors to the generic place in
<linux/pgtable.h> and then simply drop the respective definitions from the
other architectures.
The architectures that didn't provide pmd_page_vaddr() are updated to have
that defined.
The generic implementation of pte_offset_kernel() can be overridden by an
architecture and alpha makes use of this because it has special ordering
requirements for its version of pte_offset_kernel().
Any suggestions what could be the problem?
Thanks,
Adrian
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@xxxxxxxxxx
`. `' Freie Universitaet Berlin - glaubitz@xxxxxxxxxxxxxxxxxxx
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913