On Tue, 19 Jan 2016, Shannon Zhao wrote: > On 2016/1/18 22:59, Stefano Stabellini wrote: > > On Fri, 15 Jan 2016, Shannon Zhao wrote: > >> From: Shannon Zhao <shannon.zhao@xxxxxxxxxx> > >> > >> Move xen_early_init() before efi_init(), then when calling efi_init() > >> could initialize Xen specific UEFI. > >> > >> Check if it runs on Xen hypervisor through the flat dts. > >> > >> Signed-off-by: Shannon Zhao <shannon.zhao@xxxxxxxxxx> > >> --- > >> arch/arm/xen/enlighten.c | 56 ++++++++++++++++++++++++++++++++++------------- > >> arch/arm64/kernel/setup.c | 2 +- > >> 2 files changed, 42 insertions(+), 16 deletions(-) > >> > >> diff --git a/arch/arm/xen/enlighten.c b/arch/arm/xen/enlighten.c > >> index 4e78c49..5d0fe68 100644 > >> --- a/arch/arm/xen/enlighten.c > >> +++ b/arch/arm/xen/enlighten.c > >> @@ -19,6 +19,7 @@ > >> #include <linux/irqreturn.h> > >> #include <linux/module.h> > >> #include <linux/of.h> > >> +#include <linux/of_fdt.h> > >> #include <linux/of_irq.h> > >> #include <linux/of_address.h> > >> #include <linux/cpuidle.h> > >> @@ -52,8 +53,6 @@ struct xen_memory_region xen_extra_mem[XEN_EXTRA_MEM_MAX_REGIONS] __initdata; > >> > >> static __read_mostly unsigned int xen_events_irq; > >> > >> -static __initdata struct device_node *xen_node; > >> - > >> int xen_remap_domain_gfn_array(struct vm_area_struct *vma, > >> unsigned long addr, > >> xen_pfn_t *gfn, int nr, > >> @@ -222,6 +221,33 @@ static irqreturn_t xen_arm_callback(int irq, void *arg) > >> return IRQ_HANDLED; > >> } > >> > >> +static __initdata struct { > >> + const char *compat; > >> + const char *prefix; > >> + const char *version; > >> + bool found; > >> +} xen_node = {"xen,xen", "xen,xen-", NULL, false}; > >> + > >> +static int __init fdt_find_xen_node(unsigned long node, const char *uname, > >> + int depth, void *data) > >> +{ > >> + const void *s = NULL; > >> + int len; > >> + > >> + if (depth != 1 || strcmp(uname, "hypervisor") != 0) > >> + return 0; > >> + > >> + if (of_flat_dt_is_compatible(node, xen_node.compat)) > >> + xen_node.found = true; > >> + > >> + s = of_get_flat_dt_prop(node, "compatible", &len); > >> + if (strlen(xen_node.prefix) + 3 < len && > >> + !strncmp(xen_node.prefix, s, strlen(xen_node.prefix))) > >> + xen_node.version = s + strlen(xen_node.prefix); > >> + > >> + return 0; > >> +} > >> + > >> /* > >> * see Documentation/devicetree/bindings/arm/xen.txt for the > >> * documentation of the Xen Device Tree format. > >> @@ -229,26 +255,18 @@ static irqreturn_t xen_arm_callback(int irq, void *arg) > >> #define GRANT_TABLE_PHYSADDR 0 > >> void __init xen_early_init(void) > >> { > >> - int len; > >> - const char *s = NULL; > >> - const char *version = NULL; > >> - const char *xen_prefix = "xen,xen-"; > >> - > >> - xen_node = of_find_compatible_node(NULL, NULL, "xen,xen"); > >> - if (!xen_node) { > >> + of_scan_flat_dt(fdt_find_xen_node, NULL); > >> + if (!xen_node.found) { > >> pr_debug("No Xen support\n"); > >> return; > >> } > >> - s = of_get_property(xen_node, "compatible", &len); > >> - if (strlen(xen_prefix) + 3 < len && > >> - !strncmp(xen_prefix, s, strlen(xen_prefix))) > >> - version = s + strlen(xen_prefix); > >> - if (version == NULL) { > >> + > >> + if (xen_node.version == NULL) { > >> pr_debug("Xen version not found\n"); > >> return; > >> } > > > > This is going to access data within the fdt, after it has been > > unflattened (by the unflatten_device_tree call). Is that allowed? > > > > If that's not allowed, then you could copy the version string to > > xen_node.version, rather than just copying the pointer. Otherwise you > > could move the version check entirely within fdt_find_xen_node and > > remove the version field from xen_node. > > > A little confused. Since here we move xen_early_init before efi_init and > before unflatten_device_tree as well, at this moment in xen_early_init > the fdt is flattened. The xen_node.version is only set in > fdt_find_xen_node and accessed in xen_early_init. You are right, sorry, I must have been confused. It might be good to rename the xen_node variable that this patch introduced in xen_guest_init. > >> - pr_info("Xen %s support found\n", version); > >> + pr_info("Xen %s support found\n", xen_node.version); > >> > >> xen_domain_type = XEN_HVM_DOMAIN; > >> > >> @@ -296,6 +314,14 @@ static int __init xen_guest_init(void) > >> } > >> #endif > >> } else { > >> + struct device_node *xen_node; > >> + > >> + xen_node = of_find_compatible_node(NULL, NULL, "xen,xen"); > >> + if (!xen_node) { > >> + pr_debug("No Xen support\n"); > >> + return -ENODEV; > >> + } > > > > It makes sense to move this into a DT specific init function called from > > xen_guest_init. Also, given that at this point we know that a Xen node > > should be present, it makes sense to return EFAULT on error, instead of > > ENODEV. The error message could be changed to something like: "Xen > > support was detected before, but it has disappeared" and should be > > pr_err. > > > OK, thanks. -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html