| Message ID | 20210308012413.14383-4-mark.tomlinson@alliedtelesis.co.nz |
|---|---|
| State | Accepted |
| Delegated to: | Pablo Neira |
| Headers | show |
| Series | Don't use RCU for x_tables synchronization | expand |
Mark Tomlinson <mark.tomlinson@alliedtelesis.co.nz> wrote: > When a new table value was assigned, it was followed by a write memory > barrier. This ensured that all writes before this point would complete > before any writes after this point. However, to determine whether the > rules are unused, the sequence counter is read. To ensure that all > writes have been done before these reads, a full memory barrier is > needed, not just a write memory barrier. The same argument applies when > incrementing the counter, before the rules are read. > > Changing to using smp_mb() instead of smp_wmb() fixes the kernel panic > reported in cc00bcaa5899 (which is still present), while still > maintaining the same speed of replacing tables. > > The smb_mb() barriers potentially slow the packet path, however testing > has shown no measurable change in performance on a 4-core MIPS64 > platform. Okay, thanks for testing. I have no further feedback.
diff --git a/include/linux/netfilter/x_tables.h b/include/linux/netfilter/x_tables.h index 5deb099d156d..8ec48466410a 100644 --- a/include/linux/netfilter/x_tables.h +++ b/include/linux/netfilter/x_tables.h @@ -376,7 +376,7 @@ static inline unsigned int xt_write_recseq_begin(void) * since addend is most likely 1 */ __this_cpu_add(xt_recseq.sequence, addend); - smp_wmb(); + smp_mb(); return addend; } diff --git a/net/netfilter/x_tables.c b/net/netfilter/x_tables.c index af22dbe85e2c..a2b50596b87e 100644 --- a/net/netfilter/x_tables.c +++ b/net/netfilter/x_tables.c @@ -1387,7 +1387,7 @@ xt_replace_table(struct xt_table *table, table->private = newinfo; /* make sure all cpus see new ->private value */ - smp_wmb(); + smp_mb(); /* * Even though table entries have now been swapped, other CPU's
When a new table value was assigned, it was followed by a write memory barrier. This ensured that all writes before this point would complete before any writes after this point. However, to determine whether the rules are unused, the sequence counter is read. To ensure that all writes have been done before these reads, a full memory barrier is needed, not just a write memory barrier. The same argument applies when incrementing the counter, before the rules are read. Changing to using smp_mb() instead of smp_wmb() fixes the kernel panic reported in cc00bcaa5899 (which is still present), while still maintaining the same speed of replacing tables. The smb_mb() barriers potentially slow the packet path, however testing has shown no measurable change in performance on a 4-core MIPS64 platform. Fixes: 7f5c6d4f665b ("netfilter: get rid of atomic ops in fast path") Signed-off-by: Mark Tomlinson <mark.tomlinson@alliedtelesis.co.nz> --- include/linux/netfilter/x_tables.h | 2 +- net/netfilter/x_tables.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)