Differential D4544 Diff 12001 sys/dev/ntb/ntb_hw/ntb_hw.c

Changeset View

Standalone View

sys/dev/ntb/ntb_hw/ntb_hw.c

	Show All 40 Lines
	* Ivytown NTB BAR sizes are misreported by the hardware due to a			* Ivytown NTB BAR sizes are misreported by the hardware due to a
	* hardware issue. To work around this, query the size it should be			* hardware issue. To work around this, query the size it should be
	* configured to by the device and modify the resource to correspond to			* configured to by the device and modify the resource to correspond to
	* this new size. The BIOS on systems with this problem is required to			* this new size. The BIOS on systems with this problem is required to
	* provide enough address space to allow the driver to make this change			* provide enough address space to allow the driver to make this change
	* safely.			* safely.
	*			*
	* Ideally I could have just specified the size when I allocated the			* Ideally I could have just specified the size when I allocated the
	* resource like:			* resource like:
	* bus_alloc_resource(ntb->device,			* bus_alloc_resource(ntb->device,
	* SYS_RES_MEMORY, &bar->pci_resource_id, 0ul, ~0ul,			* SYS_RES_MEMORY, &bar->pci_resource_id, 0, ~0,
				cemUnsubmitted Not Done Inline Actions This change seems gratuitous (and wrong). Also, this diff doesn't appear to have any context? Next time use 'arc diff' or 'diff -U99999' please. cem: This change seems gratuitous (and wrong). Also, this diff doesn't appear to have any context?
				jhibbitsAuthorUnsubmitted Not Done Inline Actions Regarding the diff: I'll remember to use 'diff -U99999' in the future. I created this diff on a machine in the cluster that doesn't have arc installed, and didn't regenerate it locally before posting. Regarding the change: It should be perfectly valid. It may be a little gratuitous, given that ntb is amd64 only (though, i386 should run on hardware with it), but it's consistent with all other changes. jhibbits: Regarding the diff: I'll remember to use 'diff -U99999' in the future. I created this diff on…
				cemUnsubmitted Not Done Inline Actions It's gratuitous because it's a comment :). cem: It's gratuitous because it's a comment :).
				jhibbitsAuthorUnsubmitted Not Done Inline Actions D'oh, yes. I can remove it if you'd like. jhibbits: D'oh, yes. I can remove it if you'd like.
				jhbUnsubmitted Not Done Inline Actions OTOH, if "0, ~0" is now the "correct" way to call bus_alloc_resource(), I wouldn't want incorrect examples in comments to throw off future readers. jhb: OTOH, if "0, ~0" is now the "correct" way to call bus_alloc_resource(), I wouldn't want…
	* 1ul << bar_size_bits, RF_ACTIVE);			* 1ul << bar_size_bits, RF_ACTIVE);
	* but the PCI driver does not honor the size in this call, so we have			* but the PCI driver does not honor the size in this call, so we have
	* to modify it after the fact.			* to modify it after the fact.
	*/			*/
	if (HAS_FEATURE(NTB_BAR_SIZE_4K)) {			if (HAS_FEATURE(NTB_BAR_SIZE_4K)) {
	if (bar->pci_resource_id == PCIR_BAR(2))			if (bar->pci_resource_id == PCIR_BAR(2))
	bar_size_bits = pci_read_config(ntb->device,			bar_size_bits = pci_read_config(ntb->device,
	XEON_PBAR23SZ_OFFSET, 1);			XEON_PBAR23SZ_OFFSET, 1);
	▲ Show 20 Lines • Show All 42 Lines • Show Last 20 Lines