|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [RFC PATCH v3 7/7] xen/arm: scmi: generate scmi dt node for DomUs
On Tue, Mar 11, 2025 at 01:16:18PM +0200, Grygorii Strashko wrote:
> diff --git a/tools/libs/ctrl/xc_domain.c b/tools/libs/ctrl/xc_domain.c
> index 2ddc3f4f426d..f4ffab2021cd 100644
> --- a/tools/libs/ctrl/xc_domain.c
> +++ b/tools/libs/ctrl/xc_domain.c
> @@ -2229,6 +2229,24 @@ out:
>
> return ret;
> }
> +
> +int xc_domain_get_sci_info(xc_interface *xch, uint32_t domid,
> + uint64_t *paddr, uint32_t *func_id)
> +{
> + struct xen_domctl domctl = {};
> +
> + memset(&domctl, 0, sizeof(domctl));
This memset() looks redundant with variable initialisation.
> + domctl.cmd = XEN_DOMCTL_get_sci_info;
> + domctl.domain = domid;
There's one way to write these first few lines more nicely:
struct xen_domctl domctl = {
.cmd = XEN_DOMCTL_get_sci_info,
.domain = domid,
};
> + if ( do_domctl(xch, &domctl) != 0 )
> + return 1;
The error return value should be -1. (Returning the value returned by
do_domctl() is also an option.)
> +
> + *paddr = domctl.u.sci_info.paddr;
> + *func_id = domctl.u.sci_info.func_id;
> + return 0;
> +}
> +
> /*
> * Local variables:
> * mode: C
> diff --git a/tools/libs/light/libxl_arm.c b/tools/libs/light/libxl_arm.c
> index cdf5edb299af..cc54abc1ea79 100644
> --- a/tools/libs/light/libxl_arm.c
> +++ b/tools/libs/light/libxl_arm.c
> @@ -9,6 +9,7 @@
> #include <libfdt.h>
> #include <assert.h>
> #include <xen/device_tree_defs.h>
> +#include <xenhypfs.h>
>
> /*
> * There is no clear requirements for the total size of Virtio MMIO region.
> @@ -640,9 +641,6 @@ static int make_optee_node(libxl__gc *gc, void *fdt)
> int res;
> LOG(DEBUG, "Creating OP-TEE node in dtb");
>
> - res = fdt_begin_node(fdt, "firmware");
> - if (res) return res;
> -
> res = fdt_begin_node(fdt, "optee");
> if (res) return res;
>
> @@ -655,9 +653,6 @@ static int make_optee_node(libxl__gc *gc, void *fdt)
> res = fdt_end_node(fdt);
> if (res) return res;
>
> - res = fdt_end_node(fdt);
> - if (res) return res;
> -
> return 0;
> }
>
> @@ -1191,10 +1186,9 @@ static int copy_node(libxl__gc *gc, void *fdt, void
> *pfdt,
> return 0;
> }
>
> -static int copy_node_by_path(libxl__gc *gc, const char *path,
> - void *fdt, void *pfdt)
> +static int get_path_nodeoff(const char *path, void *pfdt)
> {
> - int nodeoff, r;
> + int nodeoff;
> const char *name = strrchr(path, '/');
>
> if (!name)
> @@ -1214,12 +1208,277 @@ static int copy_node_by_path(libxl__gc *gc, const
> char *path,
> if (strcmp(fdt_get_name(pfdt, nodeoff, NULL), name))
> return -FDT_ERR_NOTFOUND;
>
> + return nodeoff;
> +}
> +
> +static int copy_node_by_path(libxl__gc *gc, const char *path,
> + void *fdt, void *pfdt)
> +{
> + int nodeoff, r;
> +
> + nodeoff = get_path_nodeoff(path, pfdt);
> + if (nodeoff < 0)
> + return nodeoff;
> +
> r = copy_node(gc, fdt, pfdt, nodeoff, 0);
> if (r) return r;
>
> return 0;
> }
>
> +static int map_sci_page(libxl__gc *gc, uint32_t domid, uint64_t paddr,
> + uint64_t guest_addr)
> +{
> + int ret;
> + uint64_t _paddr_pfn = paddr >> XC_PAGE_SHIFT;
> + uint64_t _guest_pfn = guest_addr >> XC_PAGE_SHIFT;
> +
> + assert(paddr && guest_addr);
> + LOG(DEBUG, "[%d] mapping sci shmem page %"PRIx64, domid, _paddr_pfn);
Use LOGD() instead, to print domid:
LOGD(DEBUG, domid, "mapping sci shmem page %"PRIx64, _paddr_pfn);
And use LOGD() instead of LOG() throughout the patch whenever `domid` is
available.
> +
> + ret = xc_domain_iomem_permission(CTX->xch, domid, _paddr_pfn, 1, 1);
> + if (ret < 0) {
> + LOG(ERROR,
> + "failed give domain access to iomem page %"PRIx64,
> + _paddr_pfn);
You can use LOGED() variant, which would also print errno in the log as
well. You can use it everything you log an error from a system function
(or xc_* one which should set errno on error).
> + return ret;
> + }
> +
> + ret = xc_domain_memory_mapping(CTX->xch, domid,
> + _guest_pfn, _paddr_pfn,
> + 1, 1);
> + if (ret < 0) {
> + LOG(ERROR,
> + "failed to map to domain iomem page %"PRIx64
> + " to guest address %"PRIx64,
> + _paddr_pfn, _guest_pfn);
> + return ret;
> + }
> +
> + return 0;
> +}
> +
> +static int scmi_dt_make_shmem_node(libxl__gc *gc, void *fdt)
> +{
> + int res;
> + char buf[64];
> +
> + snprintf(buf, sizeof(buf), "scmi-shmem@%llx", GUEST_SCI_SHMEM_BASE);
> +
> + res = fdt_begin_node(fdt, buf);
> + if (res) return res;
> +
> + res = fdt_property_compat(gc, fdt, 1, "arm,scmi-shmem");
> + if (res) return res;
> +
> + res = fdt_property_regs(gc, fdt, GUEST_ROOT_ADDRESS_CELLS,
> + GUEST_ROOT_SIZE_CELLS, 1,
> + GUEST_SCI_SHMEM_BASE, GUEST_SCI_SHMEM_SIZE);
> + if (res) return res;
> +
> + res = fdt_property_cell(fdt, "phandle", GUEST_PHANDLE_SCMI);
> + if (res) return res;
> +
> + res = fdt_end_node(fdt);
> + if (res) return res;
> +
> + return 0;
> +}
> +
> +static const char *name_from_path(const char *path)
> +{
> + return strrchr(path, '/') + 1;
This function should probably return NULL in case the chr '/' isn't
found (strrchr returns NULL) instead of 0x1.
Also, it might be worth moving this function to just after "#ifdef
ENABLE_PARTIAL_DEVICE_TREE" in the source file, seen that
get_path_nodeoff() could potentially use it as well.
> +}
> +
> +static int dt_copy_properties(libxl__gc *gc, void* fdt, void *xen_fdt,
^ void *fdt.
> + const char *full_name)
> +{
> + int propoff, nameoff, r, nodeoff;
> + const struct fdt_property *prop;
> +
> + LOG(DEBUG, "Copy properties for node: %s", full_name);
> + nodeoff = get_path_nodeoff(full_name, xen_fdt);
> + if (nodeoff < 0)
> + return -FDT_ERR_NOTFOUND;
> +
> + for (propoff = fdt_first_property_offset(xen_fdt, nodeoff);
> + propoff >= 0;
> + propoff = fdt_next_property_offset(xen_fdt, propoff)) {
> +
> + if (!(prop = fdt_get_property_by_offset(xen_fdt, propoff, NULL)))
> + return -FDT_ERR_INTERNAL;
> +
> + nameoff = fdt32_to_cpu(prop->nameoff);
> +
> + /* Skipping phandle nodes in xen device-tree */
> + if (strcmp(fdt_string(xen_fdt,nameoff), "phandle") == 0 ||
> + strcmp(fdt_string(xen_fdt, nameoff), "linux,phandle") == 0)
> + continue;
> +
> + r = fdt_property(fdt, fdt_string(xen_fdt, nameoff),
> + prop->data, fdt32_to_cpu(prop->len));
> + if (r) return r;
> + }
> +
> + return (propoff != -FDT_ERR_NOTFOUND)? propoff : 0;
> +}
> +
> +static int scmi_dt_scan_node(libxl__gc *gc, void *fdt, void *pfdt,
> + void *xen_fdt, int nodeoff)
> +{
> + int rc;
Could you call this variable `r` instead? `rc` is reserved for libxl's
returns codes. See "CONVENTIONAL VARIABLE NAMES" in
libs/light/CODING_STYLE.
> + int node_next;
> + char full_name[128];
> + uint32_t phandle;
> +
> + node_next = fdt_first_subnode(pfdt, nodeoff);
> + while (node_next > 0)
> + {
> + LOG(ERROR,"Processing node %s",
> + fdt_get_name(pfdt, node_next, NULL));
> +
> + phandle = fdt_get_phandle(pfdt, node_next);
> +
> + rc = fdt_get_path(pfdt, node_next, full_name, sizeof(full_name));
> + if (rc) return rc;
> +
> + rc = fdt_begin_node(fdt, name_from_path(full_name));
> + if (rc) return rc;
> +
> + rc = dt_copy_properties(gc, fdt, xen_fdt, full_name);
> + if (rc) return rc;
> +
> + if (phandle) {
> + rc = fdt_property_cell(fdt, "phandle", phandle);
> + if (rc) return rc;
> + }
> +
> + rc = scmi_dt_scan_node(gc, fdt, pfdt, xen_fdt, node_next);
> + if (rc) return rc;
> +
> + rc = fdt_end_node(fdt);
> + if (rc) return rc;
> +
> + node_next = fdt_next_subnode(pfdt, node_next);
> + }
> +
> + return 0;
> +}
> +
> +static int scmi_hypfs_fdt_check(libxl__gc *gc, void *fdt)
> +{
> + int r;
> +
> + if (fdt_magic(fdt) != FDT_MAGIC) {
> + LOG(ERROR, "FDT is not a valid Flat Device Tree");
> + return ERROR_FAIL;
> + }
> +
> + r = fdt_check_header(fdt);
> + if (r) {
> + LOG(ERROR, "Failed to check the FDT (%d)", r);
> + return ERROR_FAIL;
> + }
> +
> + return r;
> +}
> +
> +static int scmi_dt_copy_subnodes(libxl__gc *gc, void *fdt, void *pfdt)
> +{
> + struct xenhypfs_handle *hdl;
> + struct xenhypfs_dirent *ent;
> + void *xen_fdt;
> + int rc, nodeoff;
> +
> + hdl = xenhypfs_open(NULL, 0);
> + if (!hdl)
> + return -EINVAL;
That value been return doesn't make sense. Also !0 is probably enough.
> +
> + xen_fdt = xenhypfs_read_raw(hdl, "/devicetree", &ent);
> + if (!xen_fdt) {
> + rc = errno;
Is doesn't seems taht anything in the functions constructind the device
tree look at return values as if the were "errno". There's maybe just
one libxl__prepare_dtb() which looks at those value as an FDT_* error
code, via the FDT() macro. Most of the other just check for !0 values.
If errno value is uesful here, you can at least log it by replacing
LOG() by LOGE() in this next line:
> + LOG(ERROR, "Unable to read hypfs entry: %d", rc);
`rc` which contain `errno` here isn't useful to log as is, instead use
LOGE() which will write strerror(errno) automatically. Also, maybe
writing which entry can't be read in the log might be useful.
[...]
Please use `r` to store return code of many of those new function
instead of a mixture of `res`, `ret`, `r`, `rc`. `rc` is reserved for
libxl error code, and are expected to by returned by libxl_*()
functions. But it seems that most functions dealing with the device tree
actually are expected to return FDT_* error code.
Thanks,
--
Anthony Perard | Vates XCP-ng Developer
XCP-ng & Xen Orchestra - Vates solutions
web: https://vates.tech
|
 |
Lists.xenproject.org is hosted with RackSpace, monitoring our |