On Mon, Jun 17, 2024 at 01:42:20PM +0200, Gerd Hoffmann wrote:
On Fri, Jun 14, 2024 at 01:05:44PM GMT, Kevin O'Connor wrote:
On Fri, Jun 14, 2024 at 12:54:28PM +0200, Gerd Hoffmann wrote:
Hi,
Be a bit more conservative, and only enable the window by default when the ram size extends beyond 60G. Due to the mmio window this translates to an effective working configuration limit of 58G/59G, depending on machine type.
Why 60GB not 64GB?
Could this change introduce a further regression - guests with 32GB of ram that need extra pci space now break?
Unlikely.
Well, there are a very large number of SeaBIOS installs. It's hard to quantify the impact to guests that need large PCI hotplug without much ram, compared to the impact to guests that need 64bit ram but can't handle 64bit pci.
Just to review - the original change (96a8d130) was made to encourage padded PCI mappings so that hotplug has a better chance of succeeding. It was thought that there would be few guests that would have >4G ram that could not also handle >4G pci mappings. Did I miss anything?
I'm leery of moving this heuristic to 64G of ram. I can understand the logic of >4G of ram indicating support for >4G pci. However, it seems strange to me that there would be guests with 50G of ram that can't handle >4G pci, but not similar guests with 70G of ram.. It also seems strange to me that only pci hotplug users with more than 64G of ram will benfit from padded mappings.
Some possible alternatives:
* We could do nothing for now and continue to gauge how much of a problem this is.
* We could revert 96a8d130, and go back to explicit qemu hints for pci hotplug users that need large mappings.
Are there other possibilities or other heuristics?
-Kevin