Re: [PATCH net-next v14 1/5] tun/tap: add IFF_BACKPRESSURE flag

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:

Pre-existing issues:
- [High] UAPI Bit Collision between IFF_NO_PI and IFF_NOFILTER causes silent failure to re-attach socket filters.
--

--- Patch [1]: [PATCH net-next v14 1/5] tun/tap: add IFF_BACKPRESSURE flag ---
commit a863b69cf20b684debe10f1f06a9f4273abcd009
Author: Simon Schippers <[email protected]>

tun/tap: add IFF_BACKPRESSURE flag

This commit adds the IFF_BACKPRESSURE flag to the TUN/TAP UAPI headers.
The flag acts as an opt-in switch for qdisc backpressure logic.

> diff --git a/include/uapi/linux/if_tun.h b/include/uapi/linux/if_tun.h
> index 79d53c7a1ebdc..a0ddc50a75342 100644
> --- a/include/uapi/linux/if_tun.h
> +++ b/include/uapi/linux/if_tun.h
> @@ -69,6 +69,10 @@
>  #define IFF_NAPI_FRAGS	0x0020
>  /* Used in TUNSETIFF to bring up tun/tap without carrier */
>  #define IFF_NO_CARRIER	0x0040
> +/* Stop the queue instead of dropping when the internal ring is full, so an
> + * attached qdisc applies backpressure instead of being bypassed.
> + */
> +#define IFF_BACKPRESSURE	0x0080
>  #define IFF_NO_PI	0x1000

[Severity: High]
This is a pre-existing issue, but I noticed that IFF_NO_PI shares the same bit
value (0x1000) as IFF_NOFILTER in this header.

Does this bit collision cause a silent failure when re-attaching socket filters
on persistent TUN/TAP devices?

When userspace creates a persistent device and sets IFF_NO_PI, and later
re-attaches to it via TUNSETIFF, tun_set_iff() in drivers/net/tun.c evaluates
the flags:

		err = tun_attach(tun, file, ifr->ifr_flags & IFF_NOFILTER,
				 ifr->ifr_flags & IFF_NAPI,
				 ifr->ifr_flags & IFF_NAPI_FRAGS, true);

Since ifr_flags & IFF_NOFILTER matches IFF_NO_PI (0x1000), it seems this would
pass skip_filter = true to tun_attach(), silently bypassing the BPF filters.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1