Re: [PATCH 1/2 v5] x86/kdump: always reserve the low 1MiB when the crashkernel option is specified

[lijiang]

å 2019å10æ24æ 19:33, lijiang åé:
> å 2019å10æ24æ 18:07, Simon Horman åé:
>> Hi Linbo,
>>
>> thanks for your patch.
>>
>> On Wed, Oct 23, 2019 at 10:19:11PM +0800, Lianbo Jiang wrote:
>>> Kdump kernel will reuse the first 640k region because the real mode
>>> trampoline has to work in this area. When the vmcore is dumped, the
>>> old memory in this area may be accessed, therefore, kernel has to
>>> copy the contents of the first 640k area to a backup region so that
>>> kdump kernel can read the old memory from the backup area of the
>>> first 640k area, which is done in the purgatory().
>>>
>>> But, the current handling of copying the first 640k area runs into
>>> problems when SME is enabled, kernel does not properly copy these
>>> old memory to the backup area in the purgatory(), thereby, kdump
>>> kernel reads out the encrypted contents, because the kdump kernel
>>> must access the first kernel's memory with the encryption bit set
>>> when SME is enabled in the first kernel. Please refer to this link:
>>>
>>> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=204793
>>>
>>> Finally, it causes the following errors, and the crash tool gets
>>> invalid pointers when parsing the vmcore.
>>>
>>> crash> kmem -s|grep -i invalid
>>> kmem: dma-kmalloc-512: slab:ffffd77680001c00 invalid freepointer:a6086ac099f0c5a4
>>> kmem: dma-kmalloc-512: slab:ffffd77680001c00 invalid freepointer:a6086ac099f0c5a4
>>> crash>
>>>
>>> To avoid the above errors, when the crashkernel option is specified,
>>> lets reserve the remaining low 1MiB memory(after reserving real mode
>>> memory) so that the allocated memory does not fall into the low 1MiB
>>> area, which makes us not to copy the first 640k content to a backup
>>> region in purgatory(). This indicates that it does not need to be
>>> included in crash dumps or used for anything except the processor
>>> trampolines that must live in the low 1MiB.
>>>
>>> Signed-off-by: Lianbo Jiang <lijiang@xxxxxxxxxx>
>>> ---
>>> BTW:I also tried to fix the above problem in purgatory(), but there
>>> are too many restricts in purgatory() context, for example: i can't
>>> allocate new memory to create the identity mapping page table for
>>> SME situation.
>>>
>>> Currently, there are two places where the first 640k area is needed,
>>> the first one is in the find_trampoline_placement(), another one is
>>> in the reserve_real_mode(), and their content doesn't matter.
>>>
>>> In addition, also need to clean all the code related to the backup
>>> region later.
>>>
>>>  arch/x86/realmode/init.c |  2 ++
>>>  include/linux/kexec.h    |  2 ++
>>>  kernel/kexec_core.c      | 13 +++++++++++++
>>>  3 files changed, 17 insertions(+)
>>>
>>> diff --git a/arch/x86/realmode/init.c b/arch/x86/realmode/init.c
>>> index 7dce39c8c034..064cc79a015d 100644
>>> --- a/arch/x86/realmode/init.c
>>> +++ b/arch/x86/realmode/init.c
>>> @@ -3,6 +3,7 @@
>>>  #include <linux/slab.h>
>>>  #include <linux/memblock.h>
>>>  #include <linux/mem_encrypt.h>
>>> +#include <linux/kexec.h>
>>>  
>>>  #include <asm/set_memory.h>
>>>  #include <asm/pgtable.h>
>>> @@ -34,6 +35,7 @@ void __init reserve_real_mode(void)
>>>  
>>>  	memblock_reserve(mem, size);
>>>  	set_real_mode_mem(mem);
>>> +	kexec_reserve_low_1MiB();
>>>  }
>>>  
>>>  static void __init setup_real_mode(void)
>>> diff --git a/include/linux/kexec.h b/include/linux/kexec.h
>>> index 1776eb2e43a4..30acf1d738bc 100644
>>> --- a/include/linux/kexec.h
>>> +++ b/include/linux/kexec.h
>>> @@ -306,6 +306,7 @@ extern void __crash_kexec(struct pt_regs *);
>>>  extern void crash_kexec(struct pt_regs *);
>>>  int kexec_should_crash(struct task_struct *);
>>>  int kexec_crash_loaded(void);
>>> +void __init kexec_reserve_low_1MiB(void);
>>>  void crash_save_cpu(struct pt_regs *regs, int cpu);
>>>  extern int kimage_crash_copy_vmcoreinfo(struct kimage *image);
>>>  
>>> @@ -397,6 +398,7 @@ static inline void __crash_kexec(struct pt_regs *regs) { }
>>>  static inline void crash_kexec(struct pt_regs *regs) { }
>>>  static inline int kexec_should_crash(struct task_struct *p) { return 0; }
>>>  static inline int kexec_crash_loaded(void) { return 0; }
>>> +static inline void __init kexec_reserve_low_1MiB(void) { }
>>>  #define kexec_in_progress false
>>>  #endif /* CONFIG_KEXEC_CORE */
>>>  
>>> diff --git a/kernel/kexec_core.c b/kernel/kexec_core.c
>>> index 15d70a90b50d..5bd89f1fee42 100644
>>> --- a/kernel/kexec_core.c
>>> +++ b/kernel/kexec_core.c
>>> @@ -37,6 +37,7 @@
>>>  #include <linux/compiler.h>
>>>  #include <linux/hugetlb.h>
>>>  #include <linux/frame.h>
>>> +#include <linux/memblock.h>
>>>  
>>>  #include <asm/page.h>
>>>  #include <asm/sections.h>
>>> @@ -70,6 +71,18 @@ struct resource crashk_low_res = {
>>>  	.desc  = IORES_DESC_CRASH_KERNEL
>>>  };
>>>  
>>> +/*
>>> + * When the crashkernel option is specified, only use the low
>>> + * 1MiB for the real mode trampoline.
>>> + */
>>> +void __init kexec_reserve_low_1MiB(void)
>>> +{
>>> +	if (strstr(boot_command_line, "crashkernel=")) {
>>
>> Could you comment on the issue of using strstr which
>> was raised by Hatayama-san in response to an earlier revision
>> of this patch?
>>
> 
> Thank you, Simon and Hatayama-san. Lets talk about it here.
> 
>> strstr() matches for example, ANYEXTRACHARACTERScrashkernel=ANYEXTRACHARACTERS.
>>
>> Is it enough to use cmdline_find_option_bool()?
>>
> 
> The cmdline_find_option_bool() will find a boolean option, but the crashkernel option
> is not a boolean option, maybe it looks odd. So, should we use the cmdline_find_option()
> better?
> 
> +#include <asm/cmdline.h>
> 
>  void __init kexec_reserve_low_1MiB(void)
>  {
> -       if (strstr(boot_command_line, "crashkernel=")) {
> +       char buffer[4];
> +
> +       if (cmdline_find_option(boot_command_line, "crashkernel=",
> +                               buffer, sizeof(buffer))) {
Maybe it is simpler as follow:

+       if (cmdline_find_option(boot_command_line, "crashkernel=",
+                               NULL, 0)) {

Any thoughts?

Thanks
Lianbo
>                 memblock_reserve(0, 1<<20);
>                 pr_info("Reserving the low 1MiB of memory for crashkernel\n");
>         }
> 
> And here, no need to parse the arguments of crashkernel(sometimes, which has a
> complicated syntax), so the size of buffer should be enough. What's your opinion?
> 
> Thanks
> Lianbo
> 
>> Thanks in advance!
>>
>>> +		memblock_reserve(0, 1<<20);
>>> +		pr_info("Reserving the low 1MiB of memory for crashkernel\n");
>>> +	}
>>> +}
>>> +
>>>  int kexec_should_crash(struct task_struct *p)
>>>  {
>>>  	/*
>>> -- 
>>> 2.17.1
>>>
>>>
>>> _______________________________________________
>>> kexec mailing list
>>> kexec@xxxxxxxxxxxxxxxxxxx
>>> http://lists.infradead.org/mailman/listinfo/kexec
>>>