On Tue, May 22, 2018 at 10:03 AM, Logan Gunthorpe <logang@xxxxxxxxxxxx> wrote: > > > On 22/05/18 10:56 AM, Dan Williams wrote: >> On Tue, May 22, 2018 at 9:42 AM, Logan Gunthorpe <logang@xxxxxxxxxxxx> wrote: >>> Hey Dan, >>> >>> On 21/05/18 06:07 PM, Dan Williams wrote: >>>> Without this change we could fail to register the teardown of >>>> devm_memremap_pages(). The likelihood of hitting this failure is tiny >>>> as small memory allocations almost always succeed. However, the impact >>>> of the failure is large given any future reconfiguration, or >>>> disable/enable, of an nvdimm namespace will fail forever as subsequent >>>> calls to devm_memremap_pages() will fail to setup the pgmap_radix >>>> since there will be stale entries for the physical address range. >>> >>> Sorry, I don't follow this. The change only seems to prevent a warning >>> from occurring in this situation. Won't pgmap_radix_release() still be >>> called regardless of whether this patch is applied? >> >> devm_add_action() does not call the release function, >> devm_add_action_or_reset() does. > > Oh, yes. Thanks I see that now. > >> Ah, true, good catch! >> >> We should manually kill in the !registered case. I think this means we >> need to pass in the custom kill routine, because for the pmem driver >> it's blk_freeze_queue_start(). > > It may be cleaner to just have the caller call the specific kill > function if devm_memremap_pages fails... As far as I can see by then it's too late, or we need to expose release details to the caller which defeats the purpose of devm semantics. > Though, I don't fully > understand how the nvdimm pmem driver cleans up the percpu counter. The dev_pagemap setup for pmem is entirely too subtle and arguably a layering violation as it reuses the block layer q_usage_counter percpu_ref. We arrange for that counter to be shutdown before the blk_cleanup_queue() does the same.
