Creating a guest with a high vCPU count (e.g., >32) fails because the guest's device tree buffer (DOMU_DTB_SIZE) overflows during creation. The FDT nodes for each vCPU quickly exhaust the 4KiB buffer, causing a guest creation failure.
Increase the buffer size to 16KiB to support guests up to the MAX_VIRT_CPUS limit (128). Signed-off-by: Oleksandr Tyshchenko <[email protected]> --- Noticed when testing the boundary conditions for dom0less guest creation on Arm64. Domain configuration: fdt mknod /chosen domU0 fdt set /chosen/domU0 compatible "xen,domain" fdt set /chosen/domU0 \#address-cells <0x2> fdt set /chosen/domU0 \#size-cells <0x2> fdt set /chosen/domU0 memory <0x0 0x10000 > fdt set /chosen/domU0 cpus <33> fdt set /chosen/domU0 vpl011 fdt mknod /chosen/domU0 module@40400000 fdt set /chosen/domU0/module@40400000 compatible "multiboot,kernel" "multiboot,module" fdt set /chosen/domU0/module@40400000 reg <0x0 0x40400000 0x0 0x16000 > fdt set /chosen/domU0/module@40400000 bootargs "console=ttyAMA0" Failure log: (XEN) Xen dom0less mode detected (XEN) *** LOADING DOMU cpus=33 memory=0x10000KB *** (XEN) Loading d1 kernel from boot module @ 0000000040400000 (XEN) Allocating mappings totalling 64MB for d1: (XEN) d1 BANK[0] 0x00000040000000-0x00000044000000 (64MB) (XEN) Device tree generation failed (-22). (XEN) (XEN) **************************************** (XEN) Panic on CPU 0: (XEN) Could not set up domain domU0 (rc = -22) (XEN) **************************************** --- --- xen/common/device-tree/dom0less-build.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/xen/common/device-tree/dom0less-build.c b/xen/common/device-tree/dom0less-build.c index 3f5b987ed8..d7d0a47b97 100644 --- a/xen/common/device-tree/dom0less-build.c +++ b/xen/common/device-tree/dom0less-build.c @@ -461,10 +461,12 @@ static int __init domain_handle_dtb_boot_module(struct domain *d, /* * The max size for DT is 2MB. However, the generated DT is small (not including - * domU passthrough DT nodes whose size we account separately), 4KB are enough - * for now, but we might have to increase it in the future. + * domU passthrough DT nodes whose size we account separately). The size is + * primarily driven by the number of vCPU nodes. The previous 4KiB buffer was + * insufficient for guests with high vCPU counts, so it has been increased + * to support up to the MAX_VIRT_CPUS limit (128). */ -#define DOMU_DTB_SIZE 4096 +#define DOMU_DTB_SIZE (4096 * 4) static int __init prepare_dtb_domU(struct domain *d, struct kernel_info *kinfo) { int addrcells, sizecells; -- 2.34.1
