From: Luke Starrett Date: Wed, 14 Dec 2022 14:27:07 +0000 (+0000) Subject: hw/intc/arm_gicv3: Fix GICD_TYPER ITLinesNumber advertisement X-Git-Tag: qemu-xen-4.18.0-rc5~388^2~18 X-Git-Url: http://xenbits.xensource.com/gitweb?a=commitdiff_plain;h=58dff8f7ea7aadbfacf4416c4988ad7f0f88a4c8;p=qemu-xen.git hw/intc/arm_gicv3: Fix GICD_TYPER ITLinesNumber advertisement The ARM GICv3 TRM describes that the ITLinesNumber field of GICD_TYPER register: "indicates the maximum SPI INTID that the GIC implementation supports" As SPI #0 is absolute IRQ #32, the max SPI INTID should have accounted for the internal 16x SGI's and 16x PPI's. However, the original GICv3 model subtracted off the SGI/PPI. Cosmetically this can be seen at OS boot (Linux) showing 32 shy of what should be there, i.e.: [ 0.000000] GICv3: 224 SPIs implemented Though in hw/arm/virt.c, the machine is configured for 256 SPI's. ARM virt machine likely doesn't have a problem with this because the upper 32 IRQ's don't actually have anything meaningful wired. But, this does become a functional issue on a custom use case which wants to make use of these IRQ's. Additionally, boot code (i.e. TF-A) will only init up to the number (blocks of 32) that it believes to actually be there. Signed-off-by: Luke Starrett Message-id: AM9P193MB168473D99B761E204E032095D40D9@AM9P193MB1684.EURP193.PROD.OUTLOOK.COM Reviewed-by: Peter Maydell Signed-off-by: Peter Maydell --- diff --git a/hw/intc/arm_gicv3_dist.c b/hw/intc/arm_gicv3_dist.c index eea0368118..d599fefcbc 100644 --- a/hw/intc/arm_gicv3_dist.c +++ b/hw/intc/arm_gicv3_dist.c @@ -390,9 +390,9 @@ static bool gicd_readl(GICv3State *s, hwaddr offset, * MBIS == 0 (message-based SPIs not supported) * SecurityExtn == 1 if security extns supported * CPUNumber == 0 since for us ARE is always 1 - * ITLinesNumber == (num external irqs / 32) - 1 + * ITLinesNumber == (((max SPI IntID + 1) / 32) - 1) */ - int itlinesnumber = ((s->num_irq - GIC_INTERNAL) / 32) - 1; + int itlinesnumber = (s->num_irq / 32) - 1; /* * SecurityExtn must be RAZ if GICD_CTLR.DS == 1, and * "security extensions not supported" always implies DS == 1,