Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Junichi Nomura <j-nomura@ce.jp.nec.com>
To:     Dave Young <dyoung@redhat.com>
CC:     "fanc.fnst@cn.fujitsu.com" <fanc.fnst@cn.fujitsu.com>,
        "x86@kernel.org" <x86@kernel.org>,
        "kexec@lists.infradead.org" <kexec@lists.infradead.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        "Borislav Petkov" <bp@alien8.de>, "bp@suse.de" <bp@suse.de>
Subject: Re: [PATCH] x86/boot: Use EFI setup data if provided
Thread-Topic: [PATCH] x86/boot: Use EFI setup data if provided
Thread-Index: AQHU4J7q4uHsPFz2IEST45iCmNXrz6YXLkgAgAO8tgCAAF1DAIAABQmAgAAHwoA=
Date:   Mon, 25 Mar 2019 06:47:16 +0000
Message-ID: <bcb5bd29-a3b8-e42d-8da7-a8501f695c21@ce.jp.nec.com>
References: <20190322110342.GA16202@jeru.linux.bs1.fc.nec.co.jp>
 <20190322152328.GD12472@zn.tnic>
 <20190325002740.GA6637@jeru.linux.bs1.fc.nec.co.jp>
 <20190325060128.GB9385@dhcp-128-65.nay.redhat.com>
 <20190325061929.GC9385@dhcp-128-65.nay.redhat.com>
In-Reply-To: <20190325061929.GC9385@dhcp-128-65.nay.redhat.com>
Accept-Language: en-US, ja-JP
Content-Language: ja-JP
Content-Type: text/plain; charset="iso-2022-jp"
Content-ID: <DF65BB6FE75B984091147D2A297E7562@gisp.nec.co.jp>
Content-Transfer-Encoding: 8BIT
MIME-Version: 1.0
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 3/25/19 3:19 PM, Dave Young wrote:
> On 03/25/19 at 02:01pm, Dave Young wrote:
>> On 03/25/19 at 12:27am, Junichi Nomura wrote:
>>> On Fri, Mar 22, 2019 at 04:23:28PM +0100, Borislav Petkov wrote:
>>>> On Fri, Mar 22, 2019 at 11:03:43AM +0000, Junichi Nomura wrote:
>>>>> Commit 3a63f70bf4c3a ("x86/boot: Early parse RSDP and save it in
>>>>> boot_params") broke kexec boot on EFI systems.  efi_get_rsdp_addr()
>>>>> in the early parsing code tries to search RSDP from EFI table but
>>>>> whose address is virtual.
>>>>>
>>>>> Since kexec(1) provides physical address of config_table via boot_params,
>>>>> efi_get_rsdp_addr() should look for setup_data in the same way as
>>>>> efi_systab_init() in arch/x86/platform/efi/efi.c does.
>>>>
>>>> If the kexec kernel should continue to use efi_systab_init() then you
>>>> should make efi_get_rsdp_addr() exit early in the kexec-ed kernel.
>>>
>>> I'm not sure which way kexec devel is going. Added kexec list.
>>> Here is the version that exits early in efi_get_rsdp_addr().
>>>
>>> [PATCH] x86/boot: Don't try to search RSDP from EFI when kexec-booted
>>>
>>> Commit 3a63f70bf4c3a ("x86/boot: Early parse RSDP and save it in
>>> boot_params") broke kexec boot on EFI systems.  efi_get_rsdp_addr()
>>> in the early parsing code tries to search RSDP from EFI table but
>>> whose address is virtual.
>>>
>>> Normally kexec(1) provides physical address of config_table via boot_params
>>> and EFI code uses that during initialization.
>>> For the early boot code, we just exit efi_get_rsdp_addr() early if the kernel
>>> is booted by kexec.
>>>
>>> Fixes: 3a63f70bf4c3a ("x86/boot: Early parse RSDP and save it in boot_params")
>>> Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
>>> Cc: Chao Fan <fanc.fnst@cn.fujitsu.com>
>>> Cc: Borislav Petkov <bp@suse.de>
>>>
>>> diff --git a/arch/x86/boot/compressed/acpi.c b/arch/x86/boot/compressed/acpi.c
>>> index 0ef4ad5..1cefc43 100644
>>> --- a/arch/x86/boot/compressed/acpi.c
>>> +++ b/arch/x86/boot/compressed/acpi.c
>>> @@ -44,6 +44,24 @@ static acpi_physical_address get_acpi_rsdp(void)
>>>  	return addr;
>>>  }
>>>  
>>> +static bool is_kexec_booted(void)
>>> +{
>>> +	struct setup_data *data;
>>> +
>>> +	/*
>>> +	 * kexec-tools provides EFI setup data so that kexec-ed kernel
>>> +	 * can find proper tables.
>>> +	 */
>>> +	data = (struct setup_data *) boot_params->hdr.setup_data;
>>> +	while (data) {
>>> +		if (data->type == SETUP_EFI)
>>> +			return true;
>>> +		data = (struct setup_data *) data->next;
>>> +	}
>>> +
>>> +	return false;
>>> +}
>>> +
>>>  /* Search EFI system tables for RSDP. */
>>>  static acpi_physical_address efi_get_rsdp_addr(void)
>>>  {
>>> @@ -57,6 +75,10 @@ static acpi_physical_address efi_get_rsdp_addr(void)
>>>  	int size, i;
>>>  	char *sig;
>>>  
>>> +	/* If the system is kexec-booted, poking EFI systab may not work. */
>>> +	if (is_kexec_booted())
>>> +		return 0;
>>> +
>>>  	ei = &boot_params->efi_info;
>>>  	sig = (char *)&ei->efi_loader_signature;
>>>  
>>>
>>> _______________________________________________
>>> kexec mailing list
>>> kexec@lists.infradead.org
>>> http://lists.infradead.org/mailman/listinfo/kexec
>>
>> Good catch, this way looks good to me.  But the function
>> is_kexec_booted can be compiled when #ifdef CONFIG_EFI
>>
>> Otherwise:
>>
>> Acked-by: Dave Young <dyoung@redhat.com>
>>
> 
> Hold on, I replied too quick.  One question is does the above patch
> passed your test?   It can workaround and skip the wrong phys addr
> issue, but the acpi early parsing still fails because efi_get_rsdp_addr
> return 0? 

The patch works for me.
Early parsing fails with the 2nd patch but it boots fine.
EFI initialization is done later without boot_params->acpi_rsdp_addr.
I think that's how v5.0 and earlier kernels work.

> If this is the case you may need go with your old patch.
> 
> I think normally people do not see this bug, because kernel will set the
> rsdp in boot_params->acpi_rsdp_addr.  Maybe you are testing with

I think it's only done for file-based kexec interface.

> different kernel versions, eg.
> 
> old kernel kexec to new kernel.
> 
> And the old kernel does not set boot_params->acpi_rsdp_addr
> 
> Is this correct?

I'm testing kexec from v5.1-rc1 to v5.1-rc1, i.e. same kernel.

-- 
Jun'ichi Nomura, NEC Corporation / NEC Solution Innovators, Ltd.