On Mon, Mar 01, 2021 at 12:55:53PM -0800, Dan Williams wrote: > On Sun, Feb 28, 2021 at 2:39 PM Dave Chinner <da...@fromorbit.com> wrote: > > > > On Sat, Feb 27, 2021 at 03:40:24PM -0800, Dan Williams wrote: > > > On Sat, Feb 27, 2021 at 2:36 PM Dave Chinner <da...@fromorbit.com> wrote: > > > > On Fri, Feb 26, 2021 at 02:41:34PM -0800, Dan Williams wrote: > > > > > On Fri, Feb 26, 2021 at 1:28 PM Dave Chinner <da...@fromorbit.com> > > > > > wrote: > > > > > > On Fri, Feb 26, 2021 at 12:59:53PM -0800, Dan Williams wrote: > > > > it points to, check if it points to the PMEM that is being removed, > > > > grab the page it points to, map that to the relevant struct page, > > > > run collect_procs() on that page, then kill the user processes that > > > > map that page. > > > > > > > > So why can't we walk the ptescheck the physical pages that they > > > > map to and if they map to a pmem page we go poison that > > > > page and that kills any user process that maps it. > > > > > > > > i.e. I can't see how unexpected pmem device unplug is any different > > > > to an MCE delivering a hwpoison event to a DAX mapped page. > > > > > > I guess the tradeoff is walking a long list of inodes vs walking a > > > large array of pages. > > > > Not really. You're assuming all a filesystem has to do is invalidate > > everything if a device goes away, and that's not true. Finding if an > > inode has a mapping that spans a specific device in a multi-device > > filesystem can be a lot more complex than that. Just walking inodes > > is easy - determining whihc inodes need invalidation is the hard > > part. > > That inode-to-device level of specificity is not needed for the same > reason that drop_caches does not need to be specific. If the wrong > page is unmapped a re-fault will bring it back, and re-fault will fail > for the pages that are successfully removed. > > > That's where ->corrupt_range() comes in - the filesystem is already > > set up to do reverse mapping from physical range to inode(s) > > offsets... > > Sure, but what is the need to get to that level of specificity with > the filesystem for something that should rarely happen in the course > of normal operation outside of a mistake?
I can't tell if we're conflating the "a bunch of your pmem went bad" case with the "all your dimms fell out of the machine" case. If, say, a single cacheline's worth of pmem goes bad on a node with 2TB of pmem, I certainly want that level of specificity. Just notify the users of the dead piece, don't flush the whole machine down the drain. > > > There's likely always more pages than inodes, but perhaps it's more > > > efficient to walk the 'struct page' array than sb->s_inodes? > > > > I really don't see you seem to be telling us that invalidation is an > > either/or choice. There's more ways to convert physical block > > address -> inode file offset and mapping index than brute force > > inode cache walks.... > > Yes, but I was trying to map it to an existing mechanism and the > internals of drop_pagecache_sb() are, in coarse terms, close to what > needs to happen here. Yes. XFS (with rmap enabled) can do all the iteration and walking in that function except for the invalidate_mapping_* call itself. The goal of this series is first to wire up a callback within both the block and pmem subsystems so that they can take notifications and reverse-map them through the storage stack until they reach an fs superblock. Once the information has reached XFS, it can use its own reverse mappings to figure out which pages of which inodes are now targetted. The future of DAX hw error handling can be that you throw the spitwad at us, and it's our problem to distill that into mm invalidation calls. XFS' reverse mapping data is indexed by storage location and isn't sharded by address_space, so (except for the DIMMs falling out), we don't need to walk the entire inode list or scan the entire mapping. Between XFS and DAX and mm, the mm already has the invalidation calls, xfs already has the distiller, and so all we need is that first bit. The current mm code doesn't fully solve the problem, nor does it need to, since it handles DRAM errors acceptably* already. * Actually, the hwpoison code should _also_ be calling ->corrupted_range when DRAM goes bad so that we can detect metadata failures and either reload the buffer or (if it was dirty) shut down. > > > > ..... > > > > > > IOWs, what needs to happen at this point is very filesystem > > > > specific. Assuming that "device unplug == filesystem dead" is not > > > > correct, nor is specifying a generic action that assumes the > > > > filesystem is dead because a device it is using went away. > > > > > > Ok, I think I set this discussion in the wrong direction implying any > > > mapping of this action to a "filesystem dead" event. It's just a "zap > > > all ptes" event and upper layers recover from there. > > > > Yes, that's exactly what ->corrupt_range() is intended for. It > > allows the filesystem to lock out access to the bad range > > and then recover the data. Or metadata, if that's where the bad > > range lands. If that recovery fails, it can then report a data > > loss/filesystem shutdown event to userspace and kill user procs that > > span the bad range... > > > > FWIW, is this notification going to occur before or after the device > > has been physically unplugged? > > Before. This will be operations that happen in the pmem driver > ->remove() callback. > > > i.e. what do we do about the > > time-of-unplug-to-time-of-invalidation window where userspace can > > still attempt to access the missing pmem though the > > not-yet-invalidated ptes? It may not be likely that people just yank > > pmem nvdimms out of machines, but with NVMe persistent memory > > spaces, there's every chance that someone pulls the wrong device... > > The physical removal aspect is only theoretical today. While the pmem > driver has a ->remove() path that's purely a software unbind > operation. That said the vulnerability window today is if a process > acquires a dax mapping, the pmem device hosting that filesystem goes > through an unbind / bind cycle, and then a new filesystem is created / > mounted. That old pte may be able to access data that is outside its > intended protection domain. > > Going forward, for buses like CXL, there will be a managed physical > remove operation via PCIE native hotplug. The flow there is that the > PCIE hotplug driver will notify the OS of a pending removal, trigger > ->remove() on the pmem driver, and then notify the technician (slot > status LED) that the card is safe to pull. Well, that's a relief. Can we cancel longterm RDMA leases now too? <duck> --D