Hi Minggui I don't think we need get_device_count_on_bridge this function. I think we could replace this logic by checking the parent of the PTM device is THIS root port. Thanks.

If we remove this line and use the high64_max_ram to calculate the ramsize, it's too much wast of memory: pgtable_add_map((uint64_t *)ppt_mmu_pml4_addr, MEM_4G, MEM_4G, high64_max_ram - MEM_4G,

We can remove this line.

Yes. It's just a sample of high memory start from 4G and current calculate logic don't depend on the 4G  size. For any high memory start address, we could get the MAX address of RAM as the top

fix bug: if PTM-CAP device, like TSN in multi-hostbridge could cause acrn-dm crash. original PTM PCI code has not handled multi-hostbridge case, and just handled hostbridge (00:00.0) case. this

From: Wen Qian <qian.wen@...> The current code would cause infinite calls to vq_getchain() because that: - error check of vq_getchain() return value is missing. - virtqueue misunderstand

From: Wen Qian <qian.wen@...> The current code would cause infinite calls to vq_getchain() because that: - error check of vq_getchain() return value is missing. - virtqueue misunderstand

Cancel this time, and we will start 2022 TCM from ww13 after Chinese New Year, draft agenda as below, thanks.WWTopicPresenterStatusChinese New Year BreakWW13Libvirt Enhancement Introduction to Replace

That's not true if high memory doesn't start from 4GB. Which logic need the top address of HPA? Anthony

OK. I will update my patch.

OK. OK. For more tips, users could refer to printing during parsing. OK. OK. I will double check it and reply on the PR comments. OK Ok.

Same comment as before regarding the use of SOS, replace by "Service VM" throughout the patch series (including in code and code comments). Change the message to "Failed to parse the MADT

Please do not use the SOS acronym anymore, use Service VM instead. Can you fix this in the entire patch serie please. I also prefer to use the following capitalization: * vcpu -> vCPU * pcpu ->

LGTM, thanks!

We could change this logic if we should support more then 3. BTW, if we don't limit the HPA settings, we could change to hpa_number + pointer and dynamic alloc this table. No, this is legacy and

"chenli.wei" <chenli.wei@...> writes: Does that mean we want to support no more than 3 explicit HPA/size settings? Do we need duplicate such fields in hpa[0] and here? -- Best

the old parameter format is like this: (quote from document) -s 4,virtio-net,<tap_name>,[mac=<xx:xx:xx>] in which, <tap_name> refers to network interface file name. and in our original design,

Today we assume the paths to the build directory do not contain the character `@` and build the sed commands on top of this assumption. However, there is no guarantee that this assumption holds. This

Today we assume a device node under /sys/devices/pciXXXX:XX/ always contain a `config` file which shows the configuration space of the PCI functions. This seems not the case for, at least,

The stream tracks the AML binary to be parsed as a list of bytes, not characters. It makes no sense to call `ord` with a byte as the input parameter. This patch also adjust the format of the dump and