On Tue, 30 Jan 2024 13:00:10 +0800
Jisheng Zhang [off-list ref] wrote:

On Sun, Jan 28, 2024 at 08:35:29PM +0100, Petr Tesarik wrote:

quoted

As explained by a comment in <linux/u64_stats_sync.h>, write side of struct
u64_stats_sync must ensure mutual exclusion, or one seqcount update could
be lost on 32-bit platforms, thus blocking readers forever. Such lockups
have been observed in real world after stmmac_xmit() on one CPU raced with
stmmac_napi_poll_tx() on another CPU.

To fix the issue without introducing a new lock, split the statics into
three parts:

1. fields updated only under the tx queue lock,
2. fields updated only during NAPI poll,
3. fields updated only from interrupt context,

Updates to fields in the first two groups are already serialized through
other locks. It is sufficient to split the existing struct u64_stats_sync
so that each group has its own.

Note that tx_set_ic_bit is updated from both contexts. Split this counter
so that each context gets its own, and calculate their sum to get the total
value in stmmac_get_ethtool_stats().

For the third group, multiple interrupts may be processed by different CPUs
at the same time, but interrupts on the same CPU will not nest. Move fields
from this group to a newly created per-cpu struct stmmac_pcpu_stats.

Fixes: 133466c3bbe1 ("net: stmmac: use per-queue 64 bit statistics where necessary")
Link: https://lore.kernel.org/netdev/Za173PhviYg-1qIn@torres.zugschlus.de/t/ (local)
Cc: stable@vger.kernel.org
Signed-off-by: Petr Tesarik <redacted>  

Thanks for the fix patch. One trivial improviement below
s/netdev_alloc_pcpu_stats/devm_netdev_alloc_pcpu_stats to simplify
error and exit code path.

Thanks for your review.

In fact, many other allocations in stmmac could be converted to devm_*.
I wanted to stay consistent with the existing code, but hey, you're
right there's no good reason for it.

Plus, I can send convert the other places with another patch.

With that:
Reviewed-by: Jisheng Zhang <jszhang@kernel.org>

PS: when I sent the above "net: stmmac: use per-queue 64 bit statistics
where necessary", I had an impression: there are too much statistics in
stmmac driver, I didn't see so many statistics in other eth drivers, is
it possible to remove some useless or not that useful statistic members?

I don't feel authorized to make the decision. But I also wonder about
some counters. For example, why is there tx_packets and tx_pkt_n? The
former is shown as RX packets by "ip stats show dev end0", the latter
is shown by as tx_pkt_n by "ethtools -S end0". The values do differ,
but I have no clue why, and if they are even expected to be different
or if it's a bug.

Petr T