NACK[J]/Cmnt: [SRU][Q/N/J:linux-azure][PATCH 0/1] azure: backport "net: hv_netvsc: reject RSS hash key programming without RX indirection table" Edit

[Tim Whisonant]

On Wed, Feb 25, 2026 at 06:42:27PM -0300, Vinicius Peixoto wrote:
> BugLink: https://bugs.launchpad.net/bugs/2142703
> 
> [Impact]
> 
> RSS configuration requires a valid RX indirection table. However, if the network device reports only a single RX queue, an indirection table will not be allocated. hv_netvsc currently does not check for this and attempting to set the RSS hash key will hang VMs on Azure.
> 
> [Fix]
> 
> Backport the following fix:
> 
> d23564955811 ("net: hv_netvsc: reject RSS hash key programming without RX indirection table")
> 
> [Test plan]
> 
> Build tested only; Azure will run reproducers on their side to validate the fix.
> 
> [Where problems could occur]
> 
> This is small, self-contained and has no side effects, but will change the behavior of the hv_netvsc module upon an invalid RSS hash key update operation, which could cause issues related to network packet reception for machines with multiple receive queues on Azure.
> 
> [Other info]
> SF #00427202
> 
> Aditya Garg (1):
>   net: hv_netvsc: reject RSS hash key programming without RX indirection
>     table
> 
>  drivers/net/hyperv/netvsc_drv.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> -- 
> 2.51.0
> 
> 

Hey Vinicius,

The Jammy build (cbd) is failing with this syntax error:

/build/jammy/drivers/net/hyperv/netvsc_drv.c: In function 'netvsc_set_rxfh':
/build/jammy/drivers/net/hyperv/netvsc_drv.c:1798:19: error: 'struct net_device_context' has no member named 'rx_table_sz'; did you mean 'rx_table'?
 1798 |         if (!ndc->rx_table_sz)
      |                   ^~~~~~~~~~~
      |                   rx_table
  CC      drivers/firmware/efi/esrt.o
  CC [M]  fs/nfs/nfs4session.o
  CC [M]  drivers/net/ethernet/mellanox/mlx4/fw.o
make[5]: *** [/build/jammy/scripts/Makefile.build:297: drivers/net/hyperv/netvsc_drv.o] Error 1
make[4]: *** [/build/jammy/scripts/Makefile.build:560: drivers/net/hyperv] Error 2
make[4]: *** Waiting for unfinished jobs....


The Noble build (cbd) also failed, but with a different problem that
isn't related to your change:

/build/noble/drivers/cpufreq/cpufreq.c: In function 'cpufreq_register_driver':
/build/noble/drivers/cpufreq/cpufreq.c:2947:21: warning: the comparison will always evaluate as 'true' for the address of 'name' will never be NULL [-Waddress]
 2947 |                 if (!driver_data->name ||
      |                     ^
In file included from /build/noble/drivers/cpufreq/cpufreq.c:18:
/build/noble/include/linux/cpufreq.h:329:25: note: 'name' declared here
  329 |         char            name[CPUFREQ_NAME_LEN];
      |                         ^~~~

cc1: some warnings being treated as errors