Re: [PATCH] x86, cma: reserve dma contiguous area after initmem_init()
From: Ingo Molnar
Date: Fri Oct 24 2014 - 13:19:23 EST
* Weijie Yang <weijie.yang@xxxxxxxxxxx> wrote:
> Fengguang Wu reported a BUG: Int 6: CR2 (null) on x86 platform in
> 0-day Linux Kernel Performance Test:
>
> [ 0.000000] BRK [0x025ee000, 0x025eefff] PGTABLE
> [ 0.000000] cma: dma_contiguous_reserve(limit 13ffe000)
> [ 0.000000] cma: dma_contiguous_reserve: reserving 31 MiB for global area
> [ 0.000000] BUG: Int 6: CR2 (null)
> [ 0.000000] EDI c0000000 ESI (null) EBP 41c11ea4 EBX 425cc101
> [ 0.000000] ESP 41c11e98 ES 0000007b DS 0000007b
> [ 0.000000] EDX 00000001 ECX (null) EAX 41cd8150
> [ 0.000000] vec 00000006 err (null) EIP 41072227 CS 00000060 flg 00210002
> [ 0.000000] Stack: 425cc150 (null) (null) 41c11ef4 41d4ee4d (null) 13ffe000 41c11ec4
> [ 0.000000] 41c2d900 (null) 13ffe000 (null) 4185793e 0000002e 410c2982 41c11f00
> [ 0.000000] 410c2df5 (null) (null) (null) 425cc150 00013efe (null) 41c11f28
> [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 3.17.0-next-20141008 #815
> [ 0.000000] 00000000 425cc101 41c11e48 41850786 41c11ea4 41d2b1db 41d95f71 00000006
> [ 0.000000] 00000000 c0000000 00000000 41c11ea4 425cc101 41c11e98 0000007b 0000007b
> [ 0.000000] 00000001 00000000 41cd8150 00000006 00000000 41072227 00000060 00210002
> [ 0.000000] Call Trace:
> [ 0.000000] [<41850786>] dump_stack+0x16/0x18
> [ 0.000000] [<41d2b1db>] early_idt_handler+0x6b/0x6b
> [ 0.000000] [<41072227>] ? __phys_addr+0x2e/0xca
> [ 0.000000] [<41d4ee4d>] cma_declare_contiguous+0x3c/0x2d7
> [ 0.000000] [<4185793e>] ? _raw_spin_unlock_irqrestore+0x59/0x91
> [ 0.000000] [<410c2982>] ? wake_up_klogd+0x8/0x33
> [ 0.000000] [<410c2df5>] ? console_unlock+0x448/0x461
> [ 0.000000] [<41d6d359>] dma_contiguous_reserve_area+0x27/0x47
> [ 0.000000] [<41d6d4d1>] dma_contiguous_reserve+0x158/0x163
> [ 0.000000] [<41d33e0f>] setup_arch+0x79b/0xc68
> [ 0.000000] [<4184c0b4>] ? printk+0x1c/0x1e
> [ 0.000000] [<41d2b7cf>] start_kernel+0x9c/0x456
> [ 0.000000] [<41d2b2ca>] i386_start_kernel+0x79/0x7d
>
> see detail: https://lkml.org/lkml/2014/10/8/708
>
> It is because dma_contiguous_reserve() is called before initmem_init() in x86,
> the variable high_memory is not initialized but accessed by __pa(high_memory)
> in dma_contiguous_reserve().
>
> This patch moves dma_contiguous_reserve() after initmem_init() so that
> high_memory is initialized before accessed.
>
> Reported-by: Fengguang Wu <fengguang.wu@xxxxxxxxx>
> Signed-off-by: Weijie Yang <weijie.yang@xxxxxxxxxxx>
> ---
> arch/x86/kernel/setup.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
> index 235cfd3..ab08aa2 100644
> --- a/arch/x86/kernel/setup.c
> +++ b/arch/x86/kernel/setup.c
> @@ -1128,7 +1128,6 @@ void __init setup_arch(char **cmdline_p)
> setup_real_mode();
>
> memblock_set_current_limit(get_max_mapped());
> - dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
>
> /*
> * NOTE: On x86-32, only from this point on, fixmaps are ready for use.
> @@ -1159,6 +1158,7 @@ void __init setup_arch(char **cmdline_p)
> early_acpi_boot_init();
>
> initmem_init();
> + dma_contiguous_reserve(max_pfn_mapped << PAGE_SHIFT);
>
> /*
> * Reserve memory for crash kernel after SRAT is parsed so that it
I think this might be a boot regression that went upstream via
akpm's tree, via the recent 'cma' change:
5ea3b1b2f8ad cma: add placement specifier for "cma=" kernel parameter
Next time around please put 'x86' into patches that
enable features on x86 and which change its early init
behavior. The 'cma: ...' prefix was rather misleading.
Thanks,
Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/