check from a (soft) interrupt handler. But if a platform does not otherwise
require the pmap_tlb_miss_lock, then where will be a brief window of
inconsistency that, while harmless, will still fire an assertion in the
consistency check.
Fix this with the following changes:
1- Refactor the pmap_tlb_miss_lock into MI code and rename it from
pmap_tlb_miss_lock_{enter,exit}() to pmap_tlb_miss_lock_{enter,exit}().
MD code can still define the "md" hooks as necessary, and if so, will
override the common implementation.
2- Provde a pmap_bootstrap_common() function to perform common pmap bootstrap
operations, namely initializing the pmap_tlb_miss_lock if it's needed.
If MD code overrides the implementation, it's responsible for initializing
its own lock.
3- Call pmap_bootstrap_common() from the mips, powerpc booke, and riscv
pmap_bootstrap() routines. (This required adding one for riscv.)
4- Switch powerpc booke to the common pmap_tlb_miss_lock.
5- Enable pmap_tlb_miss_lock if DEBUG is defined, even if it's not otherwise
required.
PR port-mips/55062 (Failed assertion in pmap_md_tlb_check_entry())
The noted argument is wrong - if it's already been deleted then the id we
have for it is invalid.
Because we don't track deletions to the ruleset, working it out is
problematic at best.
Instead, if we have already added the rule treat it as a non-op.
This is a valid use case because we might receive a burst of messages
in the downstream application for the same address and process them
one by one. It's not the job of the downstream application to track
blacklistd state.
fd -1 is invalid, so don't query it for protocol, port or address.
fd is supposed to represent how the client is connected, but if we are
parsing route(4) messages or log files then there is no client connection
to interogate.
- Put PV locking back in place with only a minor performance impact.
pmap_enter() still needs more work - it's not easy to satisfy all the
competing requirements so I'll do that with another change.
- Use pmap_find_ptp() (lookup only) in preference to pmap_get_ptp() (alloc).
Make pm_ptphint indexed by VA not PA. Replace the per-pmap radixtree for
dynamic PV entries with a per-PTP rbtree. Cuts system time during kernel
build by ~10% for me.
This enables us to load modules depended to __ffssi2.
It is difficult to deal with weak symbols consistently in in-kernel
linker. See explanation by pgoyette on tech-kern:
http://mail-index.netbsd.org/tech-kern/2020/03/09/msg026148.html
Also, we do not currently provide ffs(9) as a kernel routine.
pmap_tlb_update_addr() indicates that the VA+ASID was not found in the TLB.
It's a harmless race condition that can happen for legitimate reasons (e.g.
a TLB miss in an interrupt handler that evicts the entry from the TLB).
See discussion:
http://mail-index.netbsd.org/port-mips/2020/03/07/msg000927.html
accordingly and structure it like other NetBSD drivers so as to re-use
the original Rx buffer rather than doing a needless free/alloc cycle.
Note this happened to work previously on my Qube2 because IP, TCP, etc.
perform their own alignment checks and react accordingly. However, it's
not clear that ALL protocols do this yet, so it's better to just do the
safe thing for now.
