Add support for CUDA forward compatibility

3XX0 · 3XX0 · commit 35a9f27c0200 · 2018-09-13T16:48:00.000-07:00
diff --git a/src/driver.c b/src/driver.c
@@ -29,9 +29,6 @@
 #include "utils.h"
 #include "xfuncs.h"
 
-#define SONAME_LIBCUDA "libcuda.so.1"
-#define SONAME_LIBNVML "libnvidia-ml.so.1"
-
 #define MAX_DEVICES     64
 #define REAP_TIMEOUT_MS 10
 
diff --git a/src/driver.h b/src/driver.h
@@ -18,6 +18,9 @@ SVCXPRT *svcunixfd_create(int, u_int, u_int);
 
 #include "error.h"
 
+#define SONAME_LIBCUDA "libcuda.so.1"
+#define SONAME_LIBNVML "libnvidia-ml.so.1"
+
 #define SOCK_CLT 0
 #define SOCK_SVC 1
 
diff --git a/src/nvc.h b/src/nvc.h
@@ -77,6 +77,7 @@ struct nvc_container_config {
         char *bins_dir;
         char *libs_dir;
         char *libs32_dir;
+        char *cudart_dir;
         char *ldconfig;
 };
 
diff --git a/src/nvc_container.c b/src/nvc_container.c
@@ -26,6 +26,7 @@ static char *cgroup_root(char *, char *, const char *);
 static char *parse_proc_file(struct error *, const char *, parse_fn, char *, const char *);
 static char *find_cgroup_path(struct error *, const struct nvc_container *, const char *);
 static char *find_namespace_path(struct error *, const struct nvc_container *, const char *);
+static int  find_library_paths(struct error *, struct nvc_container *);
 static int  lookup_owner(struct error *, struct nvc_container *);
 static int  copy_config(struct error *, struct nvc_container *, const struct nvc_container_config *);
 
@@ -180,6 +181,48 @@ find_namespace_path(struct error *err, const struct nvc_container *cnt, const ch
         return (ns);
 }
 
+static int
+find_library_paths(struct error *err, struct nvc_container *cnt)
+{
+        char path[PATH_MAX];
+        glob_t gl;
+        int rv = -1;
+        char **ptr;
+
+        if (!(cnt->flags & OPT_COMPUTE_LIBS))
+                return (0);
+
+        if (path_join(err, path, cnt->cfg.rootfs, cnt->cfg.cudart_dir) < 0)
+                return (-1);
+        if (path_append(err, path, "compat/lib*.so.*") < 0)
+                return (-1);
+
+        if (xglob(err, path, GLOB_ERR, NULL, &gl) < 0)
+                goto fail;
+        if (gl.gl_pathc > 0) {
+                cnt->nlibs = gl.gl_pathc;
+                cnt->libs = ptr = array_new(err, gl.gl_pathc);
+                if (cnt->libs == NULL)
+                        goto fail;
+
+                for (size_t i = 0; i < gl.gl_pathc; ++i) {
+                        if (path_resolve(err, path, cnt->cfg.rootfs, gl.gl_pathv[i] + strlen(cnt->cfg.rootfs)) < 0)
+                                goto fail;
+                        if (!str_array_match(path, (const char * const *)cnt->libs, (size_t)(ptr - cnt->libs))) {
+                                log_infof("selecting %s%s", cnt->cfg.rootfs, path);
+                                if ((*ptr++ = xstrdup(err, path)) == NULL)
+                                        goto fail;
+                        }
+                }
+                array_pack(cnt->libs, &cnt->nlibs);
+        }
+        rv = 0;
+
+ fail:
+        globfree(&gl);
+        return (rv);
+}
+
 static int
 lookup_owner(struct error *err, struct nvc_container *cnt)
 {
@@ -205,6 +248,7 @@ copy_config(struct error *err, struct nvc_container *cnt, const struct nvc_conta
         const char *bins_dir = cfg->bins_dir;
         const char *libs_dir = cfg->libs_dir;
         const char *libs32_dir = cfg->libs32_dir;
+        const char *cudart_dir = cfg->cudart_dir;
         const char *ldconfig = cfg->ldconfig;
         char *rootfs;
         int multiarch, ret;
@@ -266,6 +310,8 @@ copy_config(struct error *err, struct nvc_container *cnt, const struct nvc_conta
                         }
                 }
         }
+        if (cudart_dir == NULL)
+                cudart_dir = CUDA_RUNTIME_DIR;
         if (ldconfig == NULL) {
                 /*
                  * Some distributions have a wrapper script around ldconfig to reduce package install time.
@@ -284,6 +330,8 @@ copy_config(struct error *err, struct nvc_container *cnt, const struct nvc_conta
                 goto fail;
         if ((cnt->cfg.libs32_dir = xstrdup(err, libs32_dir)) == NULL)
                 goto fail;
+        if ((cnt->cfg.cudart_dir = xstrdup(err, cudart_dir)) == NULL)
+                goto fail;
         if ((cnt->cfg.ldconfig = xstrdup(err, ldconfig)) == NULL)
                 goto fail;
         rv = 0;
@@ -302,7 +350,7 @@ nvc_container_new(struct nvc_context *ctx, const struct nvc_container_config *cf
         if (validate_context(ctx) < 0)
                 return (NULL);
         if (validate_args(ctx, cfg != NULL && cfg->pid > 0 && cfg->rootfs != NULL && !str_empty(cfg->rootfs) && cfg->rootfs[0] == '/' &&
-            !str_empty(cfg->bins_dir) && !str_empty(cfg->libs_dir) && !str_empty(cfg->libs32_dir) && !str_empty(cfg->ldconfig)) < 0)
+            !str_empty(cfg->bins_dir) && !str_empty(cfg->libs_dir) && !str_empty(cfg->libs32_dir) && !str_empty(cfg->cudart_dir) && !str_empty(cfg->ldconfig)) < 0)
                 return (NULL);
         if (opts == NULL)
                 opts = default_container_opts;
@@ -322,6 +370,10 @@ nvc_container_new(struct nvc_context *ctx, const struct nvc_container_config *cf
                 goto fail;
         if (lookup_owner(&ctx->err, cnt) < 0)
                 goto fail;
+        if (!(flags & OPT_NO_CNTLIBS)) {
+                if (find_library_paths(&ctx->err, cnt) < 0)
+                        goto fail;
+        }
         if ((cnt->mnt_ns = find_namespace_path(&ctx->err, cnt, "mnt")) == NULL)
                 goto fail;
         if (!(flags & OPT_NO_CGROUPS)) {
@@ -335,6 +387,7 @@ nvc_container_new(struct nvc_context *ctx, const struct nvc_container_config *cf
         log_infof("setting bins directory to %s", cnt->cfg.bins_dir);
         log_infof("setting libs directory to %s", cnt->cfg.libs_dir);
         log_infof("setting libs32 directory to %s", cnt->cfg.libs32_dir);
+        log_infof("setting cudart directory to %s", cnt->cfg.cudart_dir);
         log_infof("setting ldconfig to %s%s", cnt->cfg.ldconfig, (cnt->cfg.ldconfig[0] == '@') ? " (host relative)" : "");
         log_infof("setting mount namespace to %s", cnt->mnt_ns);
         if (!(flags & OPT_NO_CGROUPS))
@@ -355,8 +408,10 @@ nvc_container_free(struct nvc_container *cnt)
         free(cnt->cfg.bins_dir);
         free(cnt->cfg.libs_dir);
         free(cnt->cfg.libs32_dir);
+        free(cnt->cfg.cudart_dir);
         free(cnt->cfg.ldconfig);
         free(cnt->mnt_ns);
         free(cnt->dev_cg);
+        array_free(cnt->libs, cnt->nlibs);
         free(cnt);
 }
diff --git a/src/nvc_info.c b/src/nvc_info.c
@@ -134,7 +134,7 @@ select_libraries(struct error *err, void *ptr, const char *root, const char *ori
         /* Check the driver version. */
         if ((rv = str_has_suffix(lib, info->nvrm_version)) == false)
                 goto done;
-        if (str_array_match(lib, graphics_libs_compat, nitems(graphics_libs_compat))) {
+        if (str_array_match_prefix(lib, graphics_libs_compat, nitems(graphics_libs_compat))) {
                 /* Only choose OpenGL/EGL libraries issued by NVIDIA. */
                 if ((rv = elftool_has_dependency(&et, "libnvidia-glcore.so")) != false)
                         goto done;
@@ -394,25 +394,25 @@ lookup_ipcs(struct error *err, struct nvc_driver_info *info, const char *root, i
 bool
 match_binary_flags(const char *bin, int32_t flags)
 {
-        if ((flags & OPT_UTILITY_BINS) && str_array_match(bin, utility_bins, nitems(utility_bins)))
+        if ((flags & OPT_UTILITY_BINS) && str_array_match_prefix(bin, utility_bins, nitems(utility_bins)))
                 return (true);
-        if ((flags & OPT_COMPUTE_BINS) && str_array_match(bin, compute_bins, nitems(compute_bins)))
+        if ((flags & OPT_COMPUTE_BINS) && str_array_match_prefix(bin, compute_bins, nitems(compute_bins)))
                 return (true);
         return (false);
 }
 
 bool
 match_library_flags(const char *lib, int32_t flags)
 {
-        if ((flags & OPT_UTILITY_LIBS) && str_array_match(lib, utility_libs, nitems(utility_libs)))
+        if ((flags & OPT_UTILITY_LIBS) && str_array_match_prefix(lib, utility_libs, nitems(utility_libs)))
                 return (true);
-        if ((flags & OPT_COMPUTE_LIBS) && str_array_match(lib, compute_libs, nitems(compute_libs)))
+        if ((flags & OPT_COMPUTE_LIBS) && str_array_match_prefix(lib, compute_libs, nitems(compute_libs)))
                 return (true);
-        if ((flags & OPT_VIDEO_LIBS) && str_array_match(lib, video_libs, nitems(video_libs)))
+        if ((flags & OPT_VIDEO_LIBS) && str_array_match_prefix(lib, video_libs, nitems(video_libs)))
                 return (true);
-        if ((flags & OPT_GRAPHICS_LIBS) && (str_array_match(lib, graphics_libs, nitems(graphics_libs)) ||
-            str_array_match(lib, graphics_libs_glvnd, nitems(graphics_libs_glvnd)) ||
-            str_array_match(lib, graphics_libs_compat, nitems(graphics_libs_compat))))
+        if ((flags & OPT_GRAPHICS_LIBS) && (str_array_match_prefix(lib, graphics_libs, nitems(graphics_libs)) ||
+            str_array_match_prefix(lib, graphics_libs_glvnd, nitems(graphics_libs_glvnd)) ||
+            str_array_match_prefix(lib, graphics_libs_compat, nitems(graphics_libs_compat))))
                 return (true);
         return (false);
 }
diff --git a/src/nvc_internal.h b/src/nvc_internal.h
@@ -36,6 +36,8 @@
 #define NV_UVM_PROC_DRIVER       "/proc/driver/nvidia-uvm"
 #define NV_APP_PROFILE_DIR       "/etc/nvidia/nvidia-application-profiles-rc.d"
 
+#define CUDA_RUNTIME_DIR         "/usr/local/cuda"
+
 struct nvc_context {
         bool initialized;
         struct error err;
@@ -51,6 +53,8 @@ struct nvc_container {
         gid_t gid;
         char *mnt_ns;
         char *dev_cg;
+        char **libs;
+        size_t nlibs;
 };
 
 enum {
diff --git a/src/nvc_mount.c b/src/nvc_mount.c
@@ -33,6 +33,7 @@ static void unmount(const char *);
 static int  setup_cgroup(struct error *, const char *, dev_t);
 static int  symlink_library(struct error *, const char *, const char *, const char *, uid_t, gid_t);
 static int  symlink_libraries(struct error *, const struct nvc_container *, const char * const [], size_t);
+static void filter_libraries(const struct nvc_driver_info *, char * [], size_t *);
 
 static char **
 mount_files(struct error *err, const char *root, const struct nvc_container *cnt, const char *dir, char *paths[], size_t size)
@@ -397,7 +398,7 @@ symlink_libraries(struct error *err, const struct nvc_container *cnt, const char
                 lib = basename(paths[i]);
                 if (str_has_prefix(lib, "libcuda.so")) {
                         /* XXX Many applications wrongly assume that libcuda.so exists (e.g. with dlopen). */
-                        if (symlink_library(err, paths[i], lib, "libcuda.so", cnt->uid, cnt->gid) < 0)
+                        if (symlink_library(err, paths[i], SONAME_LIBCUDA, "libcuda.so", cnt->uid, cnt->gid) < 0)
                                 return (-1);
                 } else if (str_has_prefix(lib, "libGLX_nvidia.so")) {
                         /* XXX GLVND requires this symlink for indirect GLX support. */
@@ -408,6 +409,27 @@ symlink_libraries(struct error *err, const struct nvc_container *cnt, const char
         return (0);
 }
 
+static void
+filter_libraries(const struct nvc_driver_info *info, char * paths[], size_t *size)
+{
+        char *lib, *maj;
+
+        /*
+         * XXX Filter out any library that matches the major version of RM to prevent us from
+         * running into an unsupported configurations (e.g. CUDA compat on Geforce or non-LTS drivers).
+         */
+        for (size_t i = 0; i < *size; ++i) {
+                lib = basename(paths[i]);
+                if ((maj = strstr(lib, ".so.")) != NULL) {
+                        maj += strlen(".so.");
+                        if (strncmp(info->nvrm_version, maj, strspn(maj, "0123456789")))
+                                continue;
+                }
+                paths[i] = NULL;
+        }
+        array_pack(paths, size);
+}
+
 int
 nvc_driver_mount(struct nvc_context *ctx, const struct nvc_container *cnt, const struct nvc_driver_info *info)
 {
@@ -423,20 +445,22 @@ nvc_driver_mount(struct nvc_context *ctx, const struct nvc_container *cnt, const
         if (ns_enter(&ctx->err, cnt->mnt_ns, CLONE_NEWNS) < 0)
                 return (-1);
 
-        nmnt = 2 + info->nbins + info->nlibs + info->nlibs32 + info->nipcs + info->ndevs;
+        nmnt = 2 + info->nbins + info->nlibs + cnt->nlibs + info->nlibs32 + info->nipcs + info->ndevs;
         mnt = ptr = (const char **)array_new(&ctx->err, nmnt);
         if (mnt == NULL)
                 goto fail;
 
         /* Procfs mount */
         if ((*ptr++ = mount_procfs(&ctx->err, ctx->cfg.root, cnt)) == NULL)
                 goto fail;
+
         /* Application profile mount */
         if (cnt->flags & OPT_GRAPHICS_LIBS) {
                 if ((*ptr++ = mount_app_profile(&ctx->err, cnt)) == NULL)
                         goto fail;
         }
-        /* Binary and library mounts */
+
+        /* Host binary and library mounts */
         if (info->bins != NULL && info->nbins > 0) {
                 if ((tmp = (const char **)mount_files(&ctx->err, ctx->cfg.root, cnt, cnt->cfg.bins_dir, info->bins, info->nbins)) == NULL)
                         goto fail;
@@ -457,6 +481,24 @@ nvc_driver_mount(struct nvc_context *ctx, const struct nvc_container *cnt, const
         }
         if (symlink_libraries(&ctx->err, cnt, mnt, (size_t)(ptr - mnt)) < 0)
                 goto fail;
+
+        /* Container library mounts */
+        if (cnt->libs != NULL && cnt->nlibs > 0) {
+                size_t nlibs = cnt->nlibs;
+                char **libs = array_copy(&ctx->err, (const char * const *)cnt->libs, cnt->nlibs);
+                if (libs == NULL)
+                        goto fail;
+
+                filter_libraries(info, libs, &nlibs);
+                if ((tmp = (const char **)mount_files(&ctx->err, cnt->cfg.rootfs, cnt, cnt->cfg.libs_dir, libs, nlibs)) == NULL) {
+                        free(libs);
+                        goto fail;
+                }
+                ptr = array_append(ptr, tmp, array_size(tmp));
+                free(tmp);
+                free(libs);
+        }
+
         /* IPC mounts */
         for (size_t i = 0; i < info->nipcs; ++i) {
                 /* XXX Only utility libraries require persistenced IPC, everything else is compute only. */
@@ -468,6 +510,7 @@ nvc_driver_mount(struct nvc_context *ctx, const struct nvc_container *cnt, const
                 if ((*ptr++ = mount_ipc(&ctx->err, ctx->cfg.root, cnt, info->ipcs[i])) == NULL)
                         goto fail;
         }
+
         /* Device mounts */
         for (size_t i = 0; i < info->ndevs; ++i) {
                 /* XXX Only compute libraries require specific devices (e.g. UVM). */
diff --git a/src/options.h b/src/options.h
@@ -55,17 +55,18 @@ enum {
         OPT_STANDALONE    = 1 << 1,
         OPT_NO_CGROUPS    = 1 << 2,
         OPT_NO_DEVBIND    = 1 << 3,
-        OPT_UTILITY_LIBS  = 1 << 4,
-        OPT_COMPUTE_LIBS  = 1 << 5,
-        OPT_VIDEO_LIBS    = 1 << 6,
-        OPT_GRAPHICS_LIBS = 1 << 7,
-        OPT_DISPLAY       = 1 << 8,
-        OPT_UTILITY_BINS  = 1 << 9,
-        OPT_COMPUTE_BINS  = 1 << 10,
+        OPT_NO_CNTLIBS    = 1 << 4,
+        OPT_UTILITY_LIBS  = 1 << 5,
+        OPT_COMPUTE_LIBS  = 1 << 6,
+        OPT_VIDEO_LIBS    = 1 << 7,
+        OPT_GRAPHICS_LIBS = 1 << 8,
+        OPT_DISPLAY       = 1 << 9,
+        OPT_UTILITY_BINS  = 1 << 10,
+        OPT_COMPUTE_BINS  = 1 << 11,
 #if defined(__powerpc64__) /* ppc64le doesn't support compat32. */
         OPT_COMPAT32      = 1 << 0,
 #else
-        OPT_COMPAT32      = 1 << 11,
+        OPT_COMPAT32      = 1 << 12,
 #endif /* defined(__powerpc64__) */
 };
 
@@ -74,6 +75,7 @@ static const struct option container_opts[] = {
         {"standalone", OPT_STANDALONE},
         {"no-cgroups", OPT_NO_CGROUPS},
         {"no-devbind", OPT_NO_DEVBIND},
+        {"no-cntlibs", OPT_NO_CNTLIBS},
         {"utility", OPT_UTILITY_BINS|OPT_UTILITY_LIBS},
         {"compute", OPT_COMPUTE_BINS|OPT_COMPUTE_LIBS},
         {"video", OPT_VIDEO_LIBS|OPT_COMPUTE_LIBS},
diff --git a/src/utils.c b/src/utils.c
diff --git a/src/utils.h b/src/utils.h
diff --git a/src/xfuncs.h b/src/xfuncs.h
