linux-kernel - Re: [Patch v2 1/4] mm: enable dump several reasons for __dump

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20200120085530.GB18028@richard>
Date:   Mon, 20 Jan 2020 16:55:30 +0800
From:   Wei Yang <richardw.yang@...ux.intel.com>
To:     Anshuman Khandual <anshuman.khandual@....com>
Cc:     Wei Yang <richardw.yang@...ux.intel.com>,
        akpm@...ux-foundation.org, linux-kernel@...r.kernel.org,
        linux-mm@...ck.org, rientjes@...gle.com
Subject: Re: [Patch v2 1/4] mm: enable dump several reasons for __dump_page()

On Mon, Jan 20, 2020 at 11:42:30AM +0530, Anshuman Khandual wrote:
>
>
>On 01/20/2020 08:34 AM, Wei Yang wrote:
>> This is a preparation to dump all reasons during check page.
>
>This really makes sense rather then just picking the reason from
>the last "if" statement.
>
>> 
>> Signed-off-by: Wei Yang <richardw.yang@...ux.intel.com>
>> ---
>>  include/linux/mmdebug.h |  2 +-
>>  mm/debug.c              | 11 ++++++-----
>>  mm/page_alloc.c         |  2 +-
>>  3 files changed, 8 insertions(+), 7 deletions(-)
>> 
>> diff --git a/include/linux/mmdebug.h b/include/linux/mmdebug.h
>> index 2ad72d2c8cc5..f0a612db8bae 100644
>> --- a/include/linux/mmdebug.h
>> +++ b/include/linux/mmdebug.h
>> @@ -10,7 +10,7 @@ struct vm_area_struct;
>>  struct mm_struct;
>>  
>>  extern void dump_page(struct page *page, const char *reason);
>> -extern void __dump_page(struct page *page, const char *reason);
>> +extern void __dump_page(struct page *page, int num, const char **reason);
>>  void dump_vma(const struct vm_area_struct *vma);
>>  void dump_mm(const struct mm_struct *mm);
>>  
>> diff --git a/mm/debug.c b/mm/debug.c
>> index 0461df1207cb..a8ac6f951f9f 100644
>> --- a/mm/debug.c
>> +++ b/mm/debug.c
>> @@ -42,11 +42,11 @@ const struct trace_print_flags vmaflag_names[] = {
>>  	{0, NULL}
>>  };
>>  
>> -void __dump_page(struct page *page, const char *reason)
>> +void __dump_page(struct page *page, int num, const char **reason)
>>  {
>>  	struct address_space *mapping;
>>  	bool page_poisoned = PagePoisoned(page);
>> -	int mapcount;
>> +	int mapcount, i;
>>  
>>  	/*
>>  	 * If struct page is poisoned don't access Page*() functions as that
>> @@ -97,8 +97,9 @@ void __dump_page(struct page *page, const char *reason)
>>  			sizeof(unsigned long), page,
>>  			sizeof(struct page), false);
>>  
>> -	if (reason)
>> -		pr_warn("page dumped because: %s\n", reason);
>> +	pr_warn("page dumped because:\n");
>> +	for (i = 0; i < num; i++)
>> +		pr_warn("\t%s\n", reason[i]);
>
>We should have a NR_BAD_PAGE_REASONS or something to cap this iteration
>and also check reason[i] for non-NULL before trying to print the array.
>There might be call sites like the following which will be problematic
>otherwise.
>
>split_huge_page_to_list() -> dump_page(head, NULL)
>

You are right, I missed this case.

>>  
>>  #ifdef CONFIG_MEMCG
>>  	if (!page_poisoned && page->mem_cgroup)
>
>While here, will it be better to move the above debug print block after
>mem_cgroup block instead ?
>

Not sure, let's see whether others have some idea.

>> @@ -108,7 +109,7 @@ void __dump_page(struct page *page, const char *reason)
>>  
>>  void dump_page(struct page *page, const char *reason)
>>  {
>> -	__dump_page(page, reason);
>> +	__dump_page(page, 1, &reason);
>>  	dump_page_owner(page);
>>  }
>>  EXPORT_SYMBOL(dump_page);
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index d047bf7d8fd4..0cf6218aaba7 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -638,7 +638,7 @@ static void bad_page(struct page *page, const char *reason,
>>  
>>  	pr_alert("BUG: Bad page state in process %s  pfn:%05lx\n",
>>  		current->comm, page_to_pfn(page));
>> -	__dump_page(page, reason);
>> +	__dump_page(page, 1, &reason);
>>  	bad_flags &= page->flags;
>>  	if (bad_flags)
>>  		pr_alert("bad because of flags: %#lx(%pGp)\n",
>> 
>
>Do we still need to have bad_flags ? After consolidating all reasons making
>a page bad should not we just print page->flags unconditionally each time and
>let the user decipher it instead. __dump_page() will print page->flags for
>each case (atleast after the new patch from Vlastimil). AFAICS, the only
>place currently consuming bad_flags is bad_page() which seems redundant after
>first calling __dump_page().

Hmm... I don't catch this. The work in __dump_page() seems a little different
from this one. Not sure we could remove it.

-- 
Wei Yang
Help you, Help me