Messages in this thread | | | From | (Eric W. Biederman) | Date | Thu, 10 Jan 2013 15:15:19 -0800 | Subject | Re: [PATCH v7u1 26/31] x86: Don't enable swiotlb if there is not enough ram for it |
| |
Yinghai Lu <yinghai@kernel.org> writes:
> On Wed, Jan 9, 2013 at 1:15 PM, Yinghai Lu <yinghai@kernel.org> wrote: >> On Wed, Jan 9, 2013 at 1:00 PM, Eric W. Biederman <ebiederm@xmission.com> wrote: >>> Yinghai Lu <yinghai@kernel.org> writes: >>> >>>> please check updated attached. It should address all your request. >>> >>> There is one significant bug that I can see. >>> >>> swiotlb_print_info tests no_iotlb_memory but no_iotlb_memory is set >>> after swiotlb_init_with_tlb returns. >> >> there is another swiotlb_print_info calling from >> pci_swiotlb_late_init >> >> void __init pci_swiotlb_late_init(void) >> { >> /* An IOMMU turned us off. */ >> if (!swiotlb) >> swiotlb_free(); >> else { >> printk(KERN_INFO "PCI-DMA: " >> "Using software bounce buffering for IO (SWIOTLB)\n"); >> swiotlb_print_info(); >> } >> } >> >> so we need that checking when swiotlb == 1, but actually we can not >> allocate that before. > > Eric, so the code is right to put checking in swiotlb_print_info ?
My biggest question was really why you didn't set no_iotlb sooner. But shrug I didn't see any real issue with the code except for it being silly.
Certainly since we are calling swiotlb_print_info from outside swiotlb.c the check is needed.
> I'd like to post the whole patchset again and ask HPA to put them in tip/next > to catch -v3.9 merging window.
Eric
| |