Skip to contentSkip to conversation

swiotlb=force fix

DISCUSSION Muli Ben-Yehuda 1 participant 1 message Jul 27, 2006

lore
Share & tools
lore
#0 lore raw

Hi Andi,

I was looking at your swiotlb=force fix at ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.18-rc2/2.6.18-rc2-mm1/broken-out/x86_64-mm-fix-swiotlb-force.patch.

In general I agree with the patch (and thanks for fixing it).

Review

Quoted 4 lines
(quoted diff) +1 -2
-	if (!iommu_detected && !no_iommu &&
-	    (end_pfn > MAX_DMA32_PFN || force_iommu))
+	if (!iommu_detected && !no_iommu && end_pfn > MAX_DMA32_PFN)
 	       swiotlb = 1;

what is the expected outcome if the user specifies 'iommu=force' with no HW IOMMU in the machine and swiotlb compiled in? we thought the path of least surprise is to enable swiotlb unconditionally in this case. You removed the check for force_iommu here so now this set of command line options will *not* turn on swiotlb unless there's more than 4G in the machine, which can be surprising.

Review

Quoted 7 lines
(quoted diff) +2 -0
+extern int swiotlb_force;
+
 #ifdef CONFIG_SWIOTLB
 extern int swiotlb;
 #else
 #define swiotlb 0
 #endif

Any reason not to move this into the CONFIG_SWIOTLB bit the same way 'extern int swiotlb' is declared?

Cheers,
Muli