Re: [PATCH] iommu/dma: Fix race condition during iova_domain initialization

From: Robin Murphy
Date: Wed Jun 22 2022 - 09:31:20 EST


On 2022-06-22 13:46, Joerg Roedel wrote:
Please re-send with

Robin Murphy <robin.murphy@xxxxxxx>

in Cc.

Apologies, I did spot this before, I've just been tied up with other things and dropping everything non-critical on the floor, so didn't get round to replying before it slipped my mind again.

In summary, I hate it, but mostly because the whole situation of calling iommu_probe_device off the back of driver probe is fundamentally broken. I'm still a few steps away from fixing that properly, at which point I can just as well rip all these little bodges out again. If it really does need mitigating in the meantime (i.e. this is real-world async probe, not just some contrived testcase), then I can't easily think of any cleaner hack, so,

Acked-by: Robin Murphy <robin.murphy@xxxxxxx>

(somewhat reluctantly)

Cheers,
Robin.

On Mon, May 30, 2022 at 08:07:45PM +0800, yf.wang@xxxxxxxxxxxx wrote:
From: Yunfei Wang <yf.wang@xxxxxxxxxxxx>

When many devices share the same iova domain, iommu_dma_init_domain()
may be called at the same time. The checking of iovad->start_pfn will
all get false in iommu_dma_init_domain() and both enter init_iova_domain()
to do iovad initialization.

Fix this by protecting init_iova_domain() with iommu_dma_cookie->mutex.

Exception backtrace:
rb_insert_color(param1=0xFFFFFF80CD2BDB40, param3=1) + 64
init_iova_domain() + 180
iommu_setup_dma_ops() + 260
arch_setup_dma_ops() + 132
of_dma_configure_id() + 468
platform_dma_configure() + 32
really_probe() + 1168
driver_probe_device() + 268
__device_attach_driver() + 524
__device_attach() + 524
bus_probe_device() + 64
deferred_probe_work_func() + 260
process_one_work() + 580
worker_thread() + 1076
kthread() + 332
ret_from_fork() + 16

Signed-off-by: Ning Li <ning.li@xxxxxxxxxxxx>
Signed-off-by: Yunfei Wang <yf.wang@xxxxxxxxxxxx>
---
drivers/iommu/dma-iommu.c | 17 +++++++++++++----
1 file changed, 13 insertions(+), 4 deletions(-)

diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
index 09f6e1c0f9c0..b38c5041eeab 100644
--- a/drivers/iommu/dma-iommu.c
+++ b/drivers/iommu/dma-iommu.c
@@ -63,6 +63,7 @@ struct iommu_dma_cookie {
/* Domain for flush queue callback; NULL if flush queue not in use */
struct iommu_domain *fq_domain;
+ struct mutex mutex;
};
static DEFINE_STATIC_KEY_FALSE(iommu_deferred_attach_enabled);
@@ -309,6 +310,7 @@ int iommu_get_dma_cookie(struct iommu_domain *domain)
if (!domain->iova_cookie)
return -ENOMEM;
+ mutex_init(&domain->iova_cookie->mutex);
return 0;
}
@@ -549,26 +551,33 @@ static int iommu_dma_init_domain(struct iommu_domain *domain, dma_addr_t base,
}
/* start_pfn is always nonzero for an already-initialised domain */
+ mutex_lock(&cookie->mutex);
if (iovad->start_pfn) {
if (1UL << order != iovad->granule ||
base_pfn != iovad->start_pfn) {
pr_warn("Incompatible range for DMA domain\n");
- return -EFAULT;
+ ret = -EFAULT;
+ goto done_unlock;
}
- return 0;
+ ret = 0;
+ goto done_unlock;
}
init_iova_domain(iovad, 1UL << order, base_pfn);
ret = iova_domain_init_rcaches(iovad);
if (ret)
- return ret;
+ goto done_unlock;
/* If the FQ fails we can simply fall back to strict mode */
if (domain->type == IOMMU_DOMAIN_DMA_FQ && iommu_dma_init_fq(domain))
domain->type = IOMMU_DOMAIN_DMA;
- return iova_reserve_iommu_regions(dev, domain);
+ ret = iova_reserve_iommu_regions(dev, domain);
+
+done_unlock:
+ mutex_unlock(&cookie->mutex);
+ return ret;
}
/**
--
2.18.0
_______________________________________________
iommu mailing list
iommu@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linuxfoundation.org/mailman/listinfo/iommu