Re: [PATCH net-next] selftests: drv-net: test drivers sleeping in ndo_get_stats64

On Sun, 05 Jan 2025 10:55:50 -0500 Willem de Bruijn wrote:

Two tiny comments, neither cause for respin.

Let me respin, not much work since comment changes shouldn't need a re-test..

...

+@ksft_disruptive +def procfs_downup_hammer(cfg) -> None:

• """
• Reading stats via procfs only holds the RCU lock, drivers often try
• to sleep when reading the stats, or don't protect against races.
• """
• # Max out the queues, we'll flip between max an 1

s/an/and/

...

• channels = ethnl.channels_get({'header': {'dev-index': cfg.ifindex}})
• if channels['combined-count'] == 0:

•    rx_type = 'rx'
  

• else:

•    rx_type = 'combined'
  

• cur_queue_cnt = channels[f'{rx_type}-count']
• max_queue_cnt = channels[f'{rx_type}-max']
• cmd(f"ethtool -L {cfg.ifname} {rx_type} {max_queue_cnt}")
• defer(cmd, f"ethtool -L {cfg.ifname} {rx_type} {cur_queue_cnt}")
• # Real test stats
• stats = __run_inf_loop("cat /proc/net/dev")
• defer(stats.kill)
• ipset = f"ip link set dev {cfg.ifname}"
• defer(ip, f"link set dev {cfg.ifname} up")

unimportant: could reference ipset here too, as in below.

Ha, that's what I did initially, but then running it I discovered that ip() adds the initial "ip", so we end up executing:

ip ip link set...