Re: [RFC PATCH 2/3] support for broken memory modules (BadRAM)

[Stefan Assmann <sassmann@kpanic.de>]

On 28.04.2011 17:51, Stefan Assmann wrote:
> On 04/28/2011 05:08 PM, Andi Kleen wrote:
>>> You're right, logging every page marked would be too verbose. That's why
>>> I wrapped that logging into pr_debug.
>>
>> pr_debug still floods the kernel log buffer. On large systems
>> it often already overflows.
> 
> That's a pain then, I understand.

I took a closer look at pr_debug and it seems that pr_debug gets
evaluated to a conditional branch and thus does not flood the log buffer
if not explicitly enabled. I confirmed that by dumping the log buffer.
So in the normal use-case things should be fine and if pr_debug really
is enabled it dumps a lot of data, which I hope is acceptable for
debugging purposes.

> 
>>
>>> However I kept the printk in the case of early allocated pages. The user
>>> should be notified of the attempt to mark a page that's already been
>>> allocated by the kernel itself.
>>
>> That's ok, although if you're unlucky (e.g. hit a large mem_map area)
>> it can be also very nosiy.
>>
>> It would be better if you fixed the printks to output ranges.
> 
> BadRAM patterns might often mark non-consecutive pages so outputting
> ranges could be more verbose than what we have now. I'll try to think
> of something to minimize log output.

How about the following:
static int __init badram_mark_pages(unsigned long addr, unsigned long mask)
{
	unsigned long pagecount = 0, is_reserved = 0;
[...]
	printk(KERN_INFO "BadRAM: mark 0x%lx with mask 0x%0lx\n", addr, mask);

	do {
[...]
		if (memblock_is_reserved(addr)) {
			pr_debug("BadRAM: page %lu reserved by kernel\n", pfn);
			is_reserved++;
			continue;
		}
[...]
		pr_debug("BadRAM: page %lu (addr 0x%0lx) marked bad "
			 "[total %lu]\n", pfn, addr, pagecount);
	} while (next_masked_address(&addr, mask));

	if (is_reserved)
		printk(KERN_WARNING "BadRAM: %lu page(s) already reserved and "
		       "could not be marked bad\n", is_reserved);

	return pagecount;
}

This way everything with possibly high volume log output is guarded by
pr_debug and only the summary gets printed by default. No log_buf
cluttering but also a bit harder to debug for somebody who's interested
in finding out which pages are already reserved.

  Stefan

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>