On Sat, Aug 02, 2025 at 12:55:47PM +0200, Tobias Heider wrote:
> With cludwig's recent pool_get() check fix I am seeing splassert
> warnings in my dmesg.
> 
> The full error is:
> splassert: assertwaitok: want -1 have 1
> Starting stack trace...
> assertwaitok() at pool_get+0x34
> pool_get() at m_clget+0x19c
> m_clget() at m_dup_pkt+0x168
> m_dup_pkt() at bwfm_rx+0x4c
> bwfm_rx() at bwfm_pci_msg_rx+0x3ac
> bwfm_pci_msg_rx() at bwfm_pci_ring_rx+0xb0
> bwfm_pci_ring_rx() at bwfm_pci_intr+0x18c
> bwfm_pci_intr() at aplintc_irq_handler+0x170
> aplintc_irq_handler() at arm_cpu_irq+0x44
> arm_cpu_irq() at handle_el1h_irq+0x68
> handle_el1h_irq() at cpu_idle_cycle+0x28
> --- interrupt ---
> cpu_idle_cycle() at sched_idle+0x1f4
> sched_idle() at proc_trampoline+0xc
> End of stack trace.
> 
> This is caused by a WAITOK in interrupt context.  We already
> handle m_dup_pkt() returning NULL so just turn it into a M_DONTWAIT.
> 
> ok?

ok jca@

> Index: dev/ic/bwfm.c
> ===================================================================
> RCS file: /cvs/src/sys/dev/ic/bwfm.c,v
> retrieving revision 1.111
> diff -u -p -r1.111 bwfm.c
> --- dev/ic/bwfm.c	19 Feb 2024 21:23:02 -0000	1.111
> +++ dev/ic/bwfm.c	2 Aug 2025 10:51:11 -0000
> @@ -2352,7 +2352,7 @@ bwfm_rx(struct bwfm_softc *sc, struct mb
>  	/* Remaining data is an ethernet packet, so align. */
>  	if ((mtod(m, paddr_t) & 0x3) != ETHER_ALIGN) {
>  		struct mbuf *m0;
> -		m0 = m_dup_pkt(m, ETHER_ALIGN, M_WAITOK);
> +		m0 = m_dup_pkt(m, ETHER_ALIGN, M_DONTWAIT);
>  		m_freem(m);
>  		if (m0 == NULL) {
>  			ifp->if_ierrors++;
> 

-- 
jca