On Mon, Nov 05, 2012 at 02:01:54PM -0800, Andrew Morton wrote:
> On Fri, 2 Nov 2012 12:33:13 -0400
> Naoya Horiguchi <n-horiguchi@xxxxxxxxxxxxx> wrote:
>
> > Printing out the information about which file can be affected by a
> > memory error in generic_error_remove_page() is helpful for user to
> > estimate the impact of the error.
> >
> > Changelog v2:
> > - dereference mapping->host after if (!mapping) check for robustness
> >
> > ...
> >
> > --- v3.7-rc3.orig/mm/truncate.c
> > +++ v3.7-rc3/mm/truncate.c
> > @@ -151,14 +151,20 @@ int truncate_inode_page(struct address_space *mapping, struct page *page)
> > */
> > int generic_error_remove_page(struct address_space *mapping, struct page *page)
> > {
> > + struct inode *inode;
> > +
> > if (!mapping)
> > return -EINVAL;
> > + inode = mapping->host;
> > /*
> > * Only punch for normal data pages for now.
> > * Handling other types like directories would need more auditing.
> > */
> > - if (!S_ISREG(mapping->host->i_mode))
> > + if (!S_ISREG(inode->i_mode))
> > return -EIO;
> > + pr_info("MCE %#lx: file info pgoff:%lu, inode:%lu, dev:%s\n",
> > + page_to_pfn(page), page_index(page),
> > + inode->i_ino, inode->i_sb->s_id);
> > return truncate_inode_page(mapping, page);
> > }
> > EXPORT_SYMBOL(generic_error_remove_page);
>
> A couple of things.
>
> - I worry that if a hardware error occurs, it might affect a large
> amount of memory all at the same time. For example, if a 4G memory
> block goes bad, this message will be printed a million times?
If the error on 4G memory block triggered by SRAO MCE and these 1M pages
are all pagecache pages, the answer is yes.
But I think that if it's a whole DIMM error, it should be reported by
another type of MCE than SRAO, so printing a million times seems to be
unlikely to happen.
> - hard-wiring "MCE" in here seems a bit of a layering violation?
> What right does the generic, core .error_remove_page() implementation
> have to assume that it was called because of an MCE?
OK, we need not assume that. I change "MCE " prefix to more specific
one like "Memory error ".
> Many CPU types don't eveh have such a thing?
No. At least currently, only SRAO MCE triggers memory_failure() and
it's defined only on some newest highend models of Intel CPUs.
Thanks,
Naoya
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>