Received-SPF: pass (google.com: domain of linux-kernel+bounces-18087-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
Date: Fri, 5 Jan 2024 08:49:11 -0800
From: Andrew Morton <akpm@linux-foundation.org>
To: Huacai Chen <chenhuacai@loongson.cn>
Cc: Baoquan He <bhe@redhat.com>, Vivek Goyal <vgoyal@redhat.com>, Dave Young
 <dyoung@redhat.com>, Youling Tang <tangyouling@kylinos.cn>,
 kexec@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] kdump: Defer the insertion of crashkernel resources
Message-Id: <20240105084911.b64f43b12b0c7e25436cb093@linux-foundation.org>
In-Reply-To: <20231229080213.2622204-1-chenhuacai@loongson.cn>
References: <20231229080213.2622204-1-chenhuacai@loongson.cn>
Precedence: bulk
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit

On Fri, 29 Dec 2023 16:02:13 +0800 Huacai Chen <chenhuacai@loongson.cn> wrote:

> In /proc/iomem, sub-regions should be inserted after their parent,
> otherwise the insertion of parent resource fails. But after generic
> crashkernel reservation applied, in both RISC-V and ARM64 (LoongArch
> will also use generic reservation later on), crashkernel resources are
> inserted before their parent, which causes the parent disappear in
> /proc/iomem. So we defer the insertion of crashkernel resources to an
> early_initcall().
> 
> ...
>
> --- a/kernel/crash_core.c
> +++ b/kernel/crash_core.c
> @@ -377,7 +377,6 @@ static int __init reserve_crashkernel_low(unsigned long long low_size)
>  
>  	crashk_low_res.start = low_base;
>  	crashk_low_res.end   = low_base + low_size - 1;
> -	insert_resource(&iomem_resource, &crashk_low_res);
>  #endif
>  	return 0;
>  }
> @@ -459,8 +458,19 @@ void __init reserve_crashkernel_generic(char *cmdline,
>  
>  	crashk_res.start = crash_base;
>  	crashk_res.end = crash_base + crash_size - 1;
> -	insert_resource(&iomem_resource, &crashk_res);
>  }
> +
> +static __init int insert_crashkernel_resources(void)
> +{
> +	if (crashk_res.start < crashk_res.end)
> +		insert_resource(&iomem_resource, &crashk_res);
> +
> +	if (crashk_low_res.start < crashk_low_res.end)
> +		insert_resource(&iomem_resource, &crashk_low_res);
> +
> +	return 0;
> +}
> +early_initcall(insert_crashkernel_resources);
>  #endif
>  
>  int crash_prepare_elf64_headers(struct crash_mem *mem, int need_kernel_map,

I'm thinking 

Fixes: 0ab97169aa0 ("crash_core: add generic function to do reservation").

Also, is this a regression?  Were earlier kernels OK?