linux-kernel - Re: [PATCH 20/35] irqchip/gic-v4.1: Allow direct invalidation of VLPIs

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Mon, 30 Sep 2019 10:20:05 +0100
From:   Marc Zyngier <maz@...nel.org>
To:     Zenghui Yu <yuzenghui@...wei.com>
Cc:     <kvmarm@...ts.cs.columbia.edu>, <linux-kernel@...r.kernel.org>,
        Lorenzo Pieralisi <lorenzo.pieralisi@....com>,
        Jason Cooper <jason@...edaemon.net>,
        Thomas Gleixner <tglx@...utronix.de>
Subject: Re: [PATCH 20/35] irqchip/gic-v4.1: Allow direct invalidation of  VLPIs

On 2019-09-28 03:02, Zenghui Yu wrote:
> On 2019/9/24 2:25, Marc Zyngier wrote:
>> Just like for INVALL, GICv4.1 has grown a VPE-aware INVLPI register.
>> Let's plumb it in and make use of the DirectLPI code in that case.
>> Signed-off-by: Marc Zyngier <maz@...nel.org>
>> ---
>>   drivers/irqchip/irq-gic-v3-its.c   | 19 +++++++++++++++++--
>>   include/linux/irqchip/arm-gic-v3.h |  1 +
>>   2 files changed, 18 insertions(+), 2 deletions(-)
>> diff --git a/drivers/irqchip/irq-gic-v3-its.c 
>> b/drivers/irqchip/irq-gic-v3-its.c
>> index b791c9beddf2..34595a7fcccb 100644
>> --- a/drivers/irqchip/irq-gic-v3-its.c
>> +++ b/drivers/irqchip/irq-gic-v3-its.c
>> @@ -1200,13 +1200,27 @@ static void wait_for_syncr(void __iomem 
>> *rdbase)
>>   static void direct_lpi_inv(struct irq_data *d)
>>   {
>> +	struct its_vlpi_map *map = get_vlpi_map(d);
>>   	struct its_collection *col;
>>   	void __iomem *rdbase;
>> +	u64 val;
>> +
>> +	if (map) {
>> +		struct its_device *its_dev = irq_data_get_irq_chip_data(d);
>> +
>> +		WARN_ON(!is_v4_1(its_dev->its));
>> +
>> +		val  = GICR_INVLPIR_V;
>> +		val |= FIELD_PREP(GICR_INVLPIR_VPEID, map->vpe->vpe_id);
>> +		val |= FIELD_PREP(GICR_INVLPIR_INTID, map->vintid);
>> +	} else {
>> +		val = d->hwirq;
>> +	}
>>
>>   	/* Target the redistributor this LPI is currently routed to */
>>   	col = irq_to_col(d);
>
> I think irq_to_col() may not work when GICv4.1 VLPIs are involved in.
>
> irq_to_col() gives us col_map[event] as the target redistributor,
> but the correct one for VLPIs should be 
> vlpi_maps[event]->vpe->col_idx.
> These two are not always pointing to the same physical RD.
> For example, if guest issues a MOVI against a VLPI, we will update 
> the
> corresponding vlpi_map->vpe and issue a VMOVI on ITS... but leave the
> col_map[event] unchanged.
>
> col_map[event] usually describes the physical LPI's CPU affinity, but
> when this physical LPI serves as something which the VLPI is backed 
> by,
> we take really little care of it.  Did I miss something here?

You didn't miss anything, and this is indeed another pretty bad bug.
The collection mapping is completely unused when the LPI becomes a
VLPI, and it is only the vpe->col_id that matters (which gets updated
on VMOVP).

This shows that irq_to_col() is the wrong abstraction, and what we're
interested is something that is more like 'irq to cpuid', allowing us
to directly point to the right distributor.

Please see the patch I just pushed[1], which does that.

Thanks,

         M.

[1] 
https://git.kernel.org/pub/scm/linux/kernel/git/maz/arm-platforms.git/commit/?h=irq/gic-v4.1-devel&id=aff363113eb26b6840136b69c2c7db2ea691db20
-- 
Jazz is not dead. It just smells funny...