Re: [PATCH net-next v2 4/9] hsr: Implement more robust duplicate discard for PRP

[Sebastian Andrzej Siewior <bigeasy@linutronix.de>]

On 2026-01-28 19:37:23 [+0100], Felix Maurer wrote:
> > > +	if (!block) {
> > > +		block = &node->block_buf[node->next_block];
> > > +		hsr_forget_seq_block(node, block);
> > > +
> > > +		memset(block, 0, sizeof(*block));
> > > +		block->time = jiffies;
> > > +		block->block_idx = block_idx;
> > > +
> > > +		res = xa_store(&node->seq_blocks, block_idx, block, GFP_ATOMIC);
> > > +		if (xa_is_err(res))
> >
> > Hi Felix,
> >
> > I ran Claude Code over this with review-prompts [1] and it flags
> > that in the error path above, the following is needed so that the
> > block can be re-used.
> >
> > 			block->time = 0;
> 
> I agree. It would nonetheless be reused at some point, but not setting
> time = 0 may lead to an unexpected block getting removed when it is
> reused (or at least an attempt to do so). I'll fix this in v3.

Not sure I follow. If xa_store() fails then the block is not added to
the "sequence-blocks" and it will be attempted once the next packet is
received, right?. Otherwise node->next_block is not updated at which
point this block will be added twice which sounds worse.

> Thanks,
>    Felix

Sebastian