4.9.306-rc2 review

@@ -172,11 +172,15 @@
  */
 # define CALL_NOSPEC						\
 	ANNOTATE_NOSPEC_ALTERNATIVE				\
-	ALTERNATIVE(						\
+	ALTERNATIVE_2(						\
 	ANNOTATE_RETPOLINE_SAFE					\
 	"call *%[thunk_target]\n",				\
 	"call __x86_indirect_thunk_%V[thunk_target]\n",		\
-	X86_FEATURE_RETPOLINE)
+	X86_FEATURE_RETPOLINE,					\
+	"lfence;\n"						\
+	ANNOTATE_RETPOLINE_SAFE					\
+	"call *%[thunk_target]\n",				\
+	X86_FEATURE_RETPOLINE_AMD)
 # define THUNK_TARGET(addr) [thunk_target] "r" (addr)
 
 #elif defined(CONFIG_X86_32) && defined(CONFIG_RETPOLINE)

@@ -186,7 +190,8 @@
  * here, anyway.
  */
 # define CALL_NOSPEC						\
-	ALTERNATIVE(						\
+	ANNOTATE_NOSPEC_ALTERNATIVE				\
+	ALTERNATIVE_2(						\
 	ANNOTATE_RETPOLINE_SAFE					\
 	"call *%[thunk_target]\n",				\
 	"       jmp    904f;\n"					\

@@ -201,7 +206,11 @@
 	"       ret;\n"						\
 	"       .align 16\n"					\
 	"904:	call   901b;\n",				\
-	X86_FEATURE_RETPOLINE)
+	X86_FEATURE_RETPOLINE,					\
+	"lfence;\n"						\
+	ANNOTATE_RETPOLINE_SAFE					\
+	"call *%[thunk_target]\n",				\
+	X86_FEATURE_RETPOLINE_AMD)
 
 # define THUNK_TARGET(addr) [thunk_target] "rm" (addr)
 #else /* No retpoline for C / inline asm */

@@ -164,11 +164,12 @@
 	_ASM_PTR " 999b\n\t"					\
 	".popsection\n\t"
 
-#if defined(CONFIG_X86_64) && defined(RETPOLINE)
+#ifdef CONFIG_RETPOLINE
+#ifdef CONFIG_X86_64
 
 /*
- * Since the inline asm uses the %V modifier which is only in newer GCC,
- * the 64-bit one is dependent on RETPOLINE not CONFIG_RETPOLINE.
+ * Inline asm uses the %V modifier which is only in newer GCC
+ * which is ensured when CONFIG_RETPOLINE is defined.
  */
 # define CALL_NOSPEC						\
 	ANNOTATE_NOSPEC_ALTERNATIVE				\

@@ -183,7 +184,7 @@
 	X86_FEATURE_RETPOLINE_AMD)
 # define THUNK_TARGET(addr) [thunk_target] "r" (addr)
 
-#elif defined(CONFIG_X86_32) && defined(CONFIG_RETPOLINE)
+#else /* CONFIG_X86_32 */
 /*
  * For i386 we use the original ret-equivalent retpoline, because
  * otherwise we'll run out of registers. We don't care about CET

@@ -213,6 +214,7 @@
 	X86_FEATURE_RETPOLINE_AMD)
 
 # define THUNK_TARGET(addr) [thunk_target] "rm" (addr)
+#endif
 #else /* No retpoline for C / inline asm */
 # define CALL_NOSPEC "call *%[thunk_target]\n"
 # define THUNK_TARGET(addr) [thunk_target] "rm" (addr)

@@ -812,7 +812,7 @@ static void __init spec_v2_print_cond(co
 
 static inline bool retp_compiler(void)
 {
-	return __is_defined(RETPOLINE);
+	return __is_defined(CONFIG_RETPOLINE);
 }
 
 static enum spectre_v2_mitigation_cmd __init spectre_v2_parse_cmdline(void)

@@ -223,8 +223,6 @@
 /* The Spectre V2 mitigation variants */
 enum spectre_v2_mitigation {
 	SPECTRE_V2_NONE,
-	SPECTRE_V2_RETPOLINE_MINIMAL,
-	SPECTRE_V2_RETPOLINE_MINIMAL_AMD,
 	SPECTRE_V2_RETPOLINE_GENERIC,
 	SPECTRE_V2_RETPOLINE_AMD,
 	SPECTRE_V2_IBRS_ENHANCED,

@@ -784,8 +784,6 @@ set_mode:
 
 static const char * const spectre_v2_strings[] = {
 	[SPECTRE_V2_NONE]			= "Vulnerable",
-	[SPECTRE_V2_RETPOLINE_MINIMAL]		= "Vulnerable: Minimal generic ASM retpoline",
-	[SPECTRE_V2_RETPOLINE_MINIMAL_AMD]	= "Vulnerable: Minimal AMD ASM retpoline",
 	[SPECTRE_V2_RETPOLINE_GENERIC]		= "Mitigation: Full generic retpoline",
 	[SPECTRE_V2_RETPOLINE_AMD]		= "Mitigation: Full AMD retpoline",
 	[SPECTRE_V2_IBRS_ENHANCED]		= "Mitigation: Enhanced IBRS",

@@ -810,11 +808,6 @@ static void __init spec_v2_print_cond(co
 		pr_info("%s selected on command line.\n", reason);
 }
 
-static inline bool retp_compiler(void)
-{
-	return __is_defined(CONFIG_RETPOLINE);
-}
-
 static enum spectre_v2_mitigation_cmd __init spectre_v2_parse_cmdline(void)
 {
 	enum spectre_v2_mitigation_cmd cmd = SPECTRE_V2_CMD_AUTO;

@@ -912,14 +905,12 @@ retpoline_auto:
 			pr_err("Spectre mitigation: LFENCE not serializing, switching to generic retpoline\n");
 			goto retpoline_generic;
 		}
-		mode = retp_compiler() ? SPECTRE_V2_RETPOLINE_AMD :
-					 SPECTRE_V2_RETPOLINE_MINIMAL_AMD;
+		mode = SPECTRE_V2_RETPOLINE_AMD;
 		setup_force_cpu_cap(X86_FEATURE_RETPOLINE_AMD);
 		setup_force_cpu_cap(X86_FEATURE_RETPOLINE);
 	} else {
 	retpoline_generic:
-		mode = retp_compiler() ? SPECTRE_V2_RETPOLINE_GENERIC :
-					 SPECTRE_V2_RETPOLINE_MINIMAL;
+		mode = SPECTRE_V2_RETPOLINE_GENERIC;
 		setup_force_cpu_cap(X86_FEATURE_RETPOLINE);
 	}
 

@@ -755,10 +755,12 @@ spectre_v2_user_select_mitigation(enum s
 	}
 
 	/*
-	 * If enhanced IBRS is enabled or SMT impossible, STIBP is not
+	 * If no STIBP, enhanced IBRS is enabled or SMT impossible, STIBP is not
 	 * required.
 	 */
-	if (!smt_possible || spectre_v2_enabled == SPECTRE_V2_IBRS_ENHANCED)
+	if (!boot_cpu_has(X86_FEATURE_STIBP) ||
+	    !smt_possible ||
+	    spectre_v2_enabled == SPECTRE_V2_IBRS_ENHANCED)
 		return;
 
 	/*

@@ -770,12 +772,6 @@ spectre_v2_user_select_mitigation(enum s
 	    boot_cpu_has(X86_FEATURE_AMD_STIBP_ALWAYS_ON))
 		mode = SPECTRE_V2_USER_STRICT_PREFERRED;
 
-	/*
-	 * If STIBP is not available, clear the STIBP mode.
-	 */
-	if (!boot_cpu_has(X86_FEATURE_STIBP))
-		mode = SPECTRE_V2_USER_NONE;
-
 	spectre_v2_user_stibp = mode;
 
 set_mode:

@@ -1254,7 +1250,6 @@ static int ib_prctl_set(struct task_stru
 		if (spectre_v2_user_ibpb == SPECTRE_V2_USER_NONE &&
 		    spectre_v2_user_stibp == SPECTRE_V2_USER_NONE)
 			return 0;
-
 		/*
 		 * With strict mode for both IBPB and STIBP, the instruction
 		 * code paths avoid checking this task flag and instead,

@@ -838,12 +838,6 @@ static enum spectre_v2_mitigation_cmd __
 		return SPECTRE_V2_CMD_AUTO;
 	}
 
-	if (cmd == SPECTRE_V2_CMD_RETPOLINE_AMD &&
-	    boot_cpu_data.x86_vendor != X86_VENDOR_AMD) {
-		pr_err("retpoline,amd selected but CPU is not AMD. Switching to AUTO select\n");
-		return SPECTRE_V2_CMD_AUTO;
-	}
-
 	spec_v2_print_cond(mitigation_options[i].option,
 			   mitigation_options[i].secure);
 	return cmd;

@@ -195,7 +195,7 @@
 #define X86_FEATURE_FENCE_SWAPGS_USER	( 7*32+10) /* "" LFENCE in user entry SWAPGS path */
 #define X86_FEATURE_FENCE_SWAPGS_KERNEL	( 7*32+11) /* "" LFENCE in kernel entry SWAPGS path */
 #define X86_FEATURE_RETPOLINE	( 7*32+12) /* "" Generic Retpoline mitigation for Spectre variant 2 */
-#define X86_FEATURE_RETPOLINE_AMD ( 7*32+13) /* "" AMD Retpoline mitigation for Spectre variant 2 */
+#define X86_FEATURE_RETPOLINE_LFENCE	( 7*32+13) /* "" Use LFENCE for Spectre variant 2 */
 
 #define X86_FEATURE_MSR_SPEC_CTRL ( 7*32+16) /* "" MSR SPEC_CTRL is implemented */
 #define X86_FEATURE_SSBD	( 7*32+17) /* Speculative Store Bypass Disable */

@@ -119,7 +119,7 @@
 	ANNOTATE_NOSPEC_ALTERNATIVE
 	ALTERNATIVE_2 __stringify(ANNOTATE_RETPOLINE_SAFE; jmp *\reg),	\
 		__stringify(RETPOLINE_JMP \reg), X86_FEATURE_RETPOLINE,	\
-		__stringify(lfence; ANNOTATE_RETPOLINE_SAFE; jmp *\reg), X86_FEATURE_RETPOLINE_AMD
+		__stringify(lfence; ANNOTATE_RETPOLINE_SAFE; jmp *\reg), X86_FEATURE_RETPOLINE_LFENCE
 #else
 	jmp	*\reg
 #endif

@@ -130,7 +130,7 @@
 	ANNOTATE_NOSPEC_ALTERNATIVE
 	ALTERNATIVE_2 __stringify(ANNOTATE_RETPOLINE_SAFE; call *\reg),	\
 		__stringify(RETPOLINE_CALL \reg), X86_FEATURE_RETPOLINE,\
-		__stringify(lfence; ANNOTATE_RETPOLINE_SAFE; call *\reg), X86_FEATURE_RETPOLINE_AMD
+		__stringify(lfence; ANNOTATE_RETPOLINE_SAFE; call *\reg), X86_FEATURE_RETPOLINE_LFENCE
 #else
 	call	*\reg
 #endif

@@ -181,7 +181,7 @@
 	"lfence;\n"						\
 	ANNOTATE_RETPOLINE_SAFE					\
 	"call *%[thunk_target]\n",				\
-	X86_FEATURE_RETPOLINE_AMD)
+	X86_FEATURE_RETPOLINE_LFENCE)
 # define THUNK_TARGET(addr) [thunk_target] "r" (addr)
 
 #else /* CONFIG_X86_32 */

@@ -211,7 +211,7 @@
 	"lfence;\n"						\
 	ANNOTATE_RETPOLINE_SAFE					\
 	"call *%[thunk_target]\n",				\
-	X86_FEATURE_RETPOLINE_AMD)
+	X86_FEATURE_RETPOLINE_LFENCE)
 
 # define THUNK_TARGET(addr) [thunk_target] "rm" (addr)
 #endif

@@ -223,8 +223,8 @@
 /* The Spectre V2 mitigation variants */
 enum spectre_v2_mitigation {
 	SPECTRE_V2_NONE,
-	SPECTRE_V2_RETPOLINE_GENERIC,
-	SPECTRE_V2_RETPOLINE_AMD,
+	SPECTRE_V2_RETPOLINE,
+	SPECTRE_V2_LFENCE,
 	SPECTRE_V2_IBRS_ENHANCED,
 };
 

@@ -620,7 +620,7 @@ enum spectre_v2_mitigation_cmd {
 	SPECTRE_V2_CMD_FORCE,
 	SPECTRE_V2_CMD_RETPOLINE,
 	SPECTRE_V2_CMD_RETPOLINE_GENERIC,
-	SPECTRE_V2_CMD_RETPOLINE_AMD,
+	SPECTRE_V2_CMD_RETPOLINE_LFENCE,
 };
 
 enum spectre_v2_user_cmd {

@@ -780,8 +780,8 @@ set_mode:
 
 static const char * const spectre_v2_strings[] = {
 	[SPECTRE_V2_NONE]			= "Vulnerable",
-	[SPECTRE_V2_RETPOLINE_GENERIC]		= "Mitigation: Full generic retpoline",
-	[SPECTRE_V2_RETPOLINE_AMD]		= "Mitigation: Full AMD retpoline",
+	[SPECTRE_V2_RETPOLINE]			= "Mitigation: Retpolines",
+	[SPECTRE_V2_LFENCE]			= "Mitigation: LFENCE",
 	[SPECTRE_V2_IBRS_ENHANCED]		= "Mitigation: Enhanced IBRS",
 };
 

@@ -793,7 +793,8 @@ static const struct {
 	{ "off",		SPECTRE_V2_CMD_NONE,		  false },
 	{ "on",			SPECTRE_V2_CMD_FORCE,		  true  },
 	{ "retpoline",		SPECTRE_V2_CMD_RETPOLINE,	  false },
-	{ "retpoline,amd",	SPECTRE_V2_CMD_RETPOLINE_AMD,	  false },
+	{ "retpoline,amd",	SPECTRE_V2_CMD_RETPOLINE_LFENCE,  false },
+	{ "retpoline,lfence",	SPECTRE_V2_CMD_RETPOLINE_LFENCE,  false },
 	{ "retpoline,generic",	SPECTRE_V2_CMD_RETPOLINE_GENERIC, false },
 	{ "auto",		SPECTRE_V2_CMD_AUTO,		  false },
 };

@@ -831,13 +832,19 @@ static enum spectre_v2_mitigation_cmd __
 	}
 
 	if ((cmd == SPECTRE_V2_CMD_RETPOLINE ||
-	     cmd == SPECTRE_V2_CMD_RETPOLINE_AMD ||
+	     cmd == SPECTRE_V2_CMD_RETPOLINE_LFENCE ||
 	     cmd == SPECTRE_V2_CMD_RETPOLINE_GENERIC) &&
 	    !IS_ENABLED(CONFIG_RETPOLINE)) {
 		pr_err("%s selected but not compiled in. Switching to AUTO select\n", mitigation_options[i].option);
 		return SPECTRE_V2_CMD_AUTO;
 	}
 
+	if ((cmd == SPECTRE_V2_CMD_RETPOLINE_LFENCE) &&
+	    !boot_cpu_has(X86_FEATURE_LFENCE_RDTSC)) {
+		pr_err("%s selected, but CPU doesn't have a serializing LFENCE. Switching to AUTO select\n", mitigation_options[i].option);
+		return SPECTRE_V2_CMD_AUTO;
+	}
+
 	spec_v2_print_cond(mitigation_options[i].option,
 			   mitigation_options[i].secure);
 	return cmd;

@@ -872,9 +879,9 @@ static void __init spectre_v2_select_mit
 		if (IS_ENABLED(CONFIG_RETPOLINE))
 			goto retpoline_auto;
 		break;
-	case SPECTRE_V2_CMD_RETPOLINE_AMD:
+	case SPECTRE_V2_CMD_RETPOLINE_LFENCE:
 		if (IS_ENABLED(CONFIG_RETPOLINE))
-			goto retpoline_amd;
+			goto retpoline_lfence;
 		break;
 	case SPECTRE_V2_CMD_RETPOLINE_GENERIC:
 		if (IS_ENABLED(CONFIG_RETPOLINE))

@@ -890,17 +897,17 @@ static void __init spectre_v2_select_mit
 
 retpoline_auto:
 	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
-	retpoline_amd:
+	retpoline_lfence:
 		if (!boot_cpu_has(X86_FEATURE_LFENCE_RDTSC)) {
 			pr_err("Spectre mitigation: LFENCE not serializing, switching to generic retpoline\n");
 			goto retpoline_generic;
 		}
-		mode = SPECTRE_V2_RETPOLINE_AMD;
-		setup_force_cpu_cap(X86_FEATURE_RETPOLINE_AMD);
+		mode = SPECTRE_V2_LFENCE;
+		setup_force_cpu_cap(X86_FEATURE_RETPOLINE_LFENCE);
 		setup_force_cpu_cap(X86_FEATURE_RETPOLINE);
 	} else {
 	retpoline_generic:
-		mode = SPECTRE_V2_RETPOLINE_GENERIC;
+		mode = SPECTRE_V2_RETPOLINE;
 		setup_force_cpu_cap(X86_FEATURE_RETPOLINE);
 	}
 

@@ -194,7 +194,7 @@
 #define X86_FEATURE_PROC_FEEDBACK ( 7*32+ 9) /* AMD ProcFeedbackInterface */
 
 #define X86_FEATURE_RETPOLINE	( 7*32+12) /* "" Generic Retpoline mitigation for Spectre variant 2 */
-#define X86_FEATURE_RETPOLINE_AMD ( 7*32+13) /* "" AMD Retpoline mitigation for Spectre variant 2 */
+#define X86_FEATURE_RETPOLINE_LFENCE	( 7*32+13) /* "" Use LFENCEs for Spectre variant 2 */
 
 #define X86_FEATURE_MSR_SPEC_CTRL ( 7*32+16) /* "" MSR SPEC_CTRL is implemented */
 #define X86_FEATURE_SSBD	( 7*32+17) /* Speculative Store Bypass Disable */

@@ -225,7 +225,9 @@ enum spectre_v2_mitigation {
 	SPECTRE_V2_NONE,
 	SPECTRE_V2_RETPOLINE,
 	SPECTRE_V2_LFENCE,
-	SPECTRE_V2_IBRS_ENHANCED,
+	SPECTRE_V2_EIBRS,
+	SPECTRE_V2_EIBRS_RETPOLINE,
+	SPECTRE_V2_EIBRS_LFENCE,
 };
 
 /* The indirect branch speculation control variants */

210 lines not on this page · raw message

@@ -30,6 +30,7 @@
 #include <asm/cacheflush.h>
 #include <asm/intel-family.h>
 #include <asm/e820.h>
+#include <linux/bpf.h>
 
 #include "cpu.h"
 

@@ -606,6 +607,16 @@ static inline const char *spectre_v2_mod
 static inline const char *spectre_v2_module_string(void) { return ""; }
 #endif
 
+#define SPECTRE_V2_EIBRS_EBPF_MSG "WARNING: Unprivileged eBPF is enabled with eIBRS on, data leaks possible via Spectre v2 BHB attacks!\n"
+
+#ifdef CONFIG_BPF_SYSCALL
+void unpriv_ebpf_notify(int new_state)
+{
+	if (spectre_v2_enabled == SPECTRE_V2_EIBRS && !new_state)
+		pr_err(SPECTRE_V2_EIBRS_EBPF_MSG);
+}
+#endif
+
 static inline bool match_option(const char *arg, int arglen, const char *opt)
 {
 	int len = strlen(opt);

@@ -949,6 +960,9 @@ static void __init spectre_v2_select_mit
 		break;
 	}
 
+	if (mode == SPECTRE_V2_EIBRS && unprivileged_ebpf_enabled())
+		pr_err(SPECTRE_V2_EIBRS_EBPF_MSG);
+
 	if (spectre_v2_in_eibrs_mode(mode)) {
 		/* Force it so VMEXIT will restore correctly */
 		x86_spec_ctrl_base |= SPEC_CTRL_IBRS;

@@ -1686,6 +1700,20 @@ static char *ibpb_state(void)
 	return "";
 }
 
+static ssize_t spectre_v2_show_state(char *buf)
+{
+	if (spectre_v2_enabled == SPECTRE_V2_EIBRS && unprivileged_ebpf_enabled())
+		return sprintf(buf, "Vulnerable: Unprivileged eBPF enabled\n");
+
+	return sprintf(buf, "%s%s%s%s%s%s\n",
+		       spectre_v2_strings[spectre_v2_enabled],
+		       ibpb_state(),
+		       boot_cpu_has(X86_FEATURE_USE_IBRS_FW) ? ", IBRS_FW" : "",
+		       stibp_state(),
+		       boot_cpu_has(X86_FEATURE_RSB_CTXSW) ? ", RSB filling" : "",
+		       spectre_v2_module_string());
+}
+
 static ssize_t srbds_show_state(char *buf)
 {
 	return sprintf(buf, "%s\n", srbds_strings[srbds_mitigation]);

@@ -1708,12 +1736,7 @@ static ssize_t cpu_show_common(struct de
 		return sprintf(buf, "%s\n", spectre_v1_strings[spectre_v1_mitigation]);
 
 	case X86_BUG_SPECTRE_V2:
-		return sprintf(buf, "%s%s%s%s%s%s\n", spectre_v2_strings[spectre_v2_enabled],
-			       ibpb_state(),
-			       boot_cpu_has(X86_FEATURE_USE_IBRS_FW) ? ", IBRS_FW" : "",
-			       stibp_state(),
-			       boot_cpu_has(X86_FEATURE_RSB_CTXSW) ? ", RSB filling" : "",
-			       spectre_v2_module_string());
+		return spectre_v2_show_state(buf);
 
 	case X86_BUG_SPEC_STORE_BYPASS:
 		return sprintf(buf, "%s\n", ssb_strings[ssb_mode]);

@@ -295,6 +295,11 @@ static inline void bpf_long_memcpy(void
 
 /* verify correctness of eBPF program */
 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
+
+static inline bool unprivileged_ebpf_enabled(void)
+{
+	return !sysctl_unprivileged_bpf_disabled;
+}
 #else
 static inline void bpf_register_prog_type(struct bpf_prog_type_list *tl)
 {

@@ -322,6 +327,12 @@ static inline struct bpf_prog *bpf_prog_
 {
 	return ERR_PTR(-EOPNOTSUPP);
 }
+
+static inline bool unprivileged_ebpf_enabled(void)
+{
+	return false;
+}
+
 #endif /* CONFIG_BPF_SYSCALL */
 
 /* verifier prototypes for helper functions called from eBPF programs */

@@ -222,6 +222,11 @@ static int sysrq_sysctl_handler(struct c
 #endif
 
 #ifdef CONFIG_BPF_SYSCALL
+
+void __weak unpriv_ebpf_notify(int new_state)
+{
+}
+
 static int bpf_unpriv_handler(struct ctl_table *table, int write,
                              void *buffer, size_t *lenp, loff_t *ppos)
 {

@@ -239,6 +244,9 @@ static int bpf_unpriv_handler(struct ctl
 			return -EPERM;
 		*(int *)table->data = unpriv_enable;
 	}
+
+	unpriv_ebpf_notify(unpriv_enable);
+
 	return ret;
 }
 #endif

@@ -897,14 +897,6 @@ static enum spectre_v2_mitigation __init
 		return SPECTRE_V2_NONE;
 	}
 
-	if (boot_cpu_data.x86_vendor == X86_VENDOR_AMD) {
-		if (!boot_cpu_has(X86_FEATURE_LFENCE_RDTSC)) {
-			pr_err("LFENCE not serializing, switching to generic retpoline\n");
-			return SPECTRE_V2_RETPOLINE;
-		}
-		return SPECTRE_V2_LFENCE;
-	}
-
 	return SPECTRE_V2_RETPOLINE;
 }
 

@@ -607,6 +607,7 @@ static inline const char *spectre_v2_mod
 static inline const char *spectre_v2_module_string(void) { return ""; }
 #endif
 
+#define SPECTRE_V2_LFENCE_MSG "WARNING: LFENCE mitigation is not recommended for this CPU, data leaks possible!\n"
 #define SPECTRE_V2_EIBRS_EBPF_MSG "WARNING: Unprivileged eBPF is enabled with eIBRS on, data leaks possible via Spectre v2 BHB attacks!\n"
 
 #ifdef CONFIG_BPF_SYSCALL

@@ -928,6 +929,7 @@ static void __init spectre_v2_select_mit
 		break;
 
 	case SPECTRE_V2_CMD_RETPOLINE_LFENCE:
+		pr_err(SPECTRE_V2_LFENCE_MSG);
 		mode = SPECTRE_V2_LFENCE;
 		break;
 

@@ -1694,6 +1696,9 @@ static char *ibpb_state(void)
 
 static ssize_t spectre_v2_show_state(char *buf)
 {
+	if (spectre_v2_enabled == SPECTRE_V2_LFENCE)
+		return sprintf(buf, "Vulnerable: LFENCE\n");
+
 	if (spectre_v2_enabled == SPECTRE_V2_EIBRS && unprivileged_ebpf_enabled())
 		return sprintf(buf, "Vulnerable: Unprivileged eBPF enabled\n");
 

@@ -609,12 +609,27 @@ static inline const char *spectre_v2_mod
 
 #define SPECTRE_V2_LFENCE_MSG "WARNING: LFENCE mitigation is not recommended for this CPU, data leaks possible!\n"
 #define SPECTRE_V2_EIBRS_EBPF_MSG "WARNING: Unprivileged eBPF is enabled with eIBRS on, data leaks possible via Spectre v2 BHB attacks!\n"
+#define SPECTRE_V2_EIBRS_LFENCE_EBPF_SMT_MSG "WARNING: Unprivileged eBPF is enabled with eIBRS+LFENCE mitigation and SMT, data leaks possible via Spectre v2 BHB attacks!\n"
 
 #ifdef CONFIG_BPF_SYSCALL
 void unpriv_ebpf_notify(int new_state)
 {
-	if (spectre_v2_enabled == SPECTRE_V2_EIBRS && !new_state)
+	if (new_state)
+		return;
+
+	/* Unprivileged eBPF is enabled */
+
+	switch (spectre_v2_enabled) {
+	case SPECTRE_V2_EIBRS:
 		pr_err(SPECTRE_V2_EIBRS_EBPF_MSG);
+		break;
+	case SPECTRE_V2_EIBRS_LFENCE:
+		if (sched_smt_active())
+			pr_err(SPECTRE_V2_EIBRS_LFENCE_EBPF_SMT_MSG);
+		break;
+	default:
+		break;
+	}
 }
 #endif
 

@@ -1074,6 +1089,10 @@ void arch_smt_update(void)
 {
 	mutex_lock(&spec_ctrl_mutex);
 
+	if (sched_smt_active() && unprivileged_ebpf_enabled() &&
+	    spectre_v2_enabled == SPECTRE_V2_EIBRS_LFENCE)
+		pr_warn_once(SPECTRE_V2_EIBRS_LFENCE_EBPF_SMT_MSG);
+
 	switch (spectre_v2_user_stibp) {
 	case SPECTRE_V2_USER_NONE:
 		break;

@@ -1700,7 +1719,11 @@ static ssize_t spectre_v2_show_state(cha
 		return sprintf(buf, "Vulnerable: LFENCE\n");
 
 	if (spectre_v2_enabled == SPECTRE_V2_EIBRS && unprivileged_ebpf_enabled())
-		return sprintf(buf, "Vulnerable: Unprivileged eBPF enabled\n");
+		return sprintf(buf, "Vulnerable: eIBRS with unprivileged eBPF\n");
+
+	if (sched_smt_active() && unprivileged_ebpf_enabled() &&
+	    spectre_v2_enabled == SPECTRE_V2_EIBRS_LFENCE)
+		return sprintf(buf, "Vulnerable: eIBRS+LFENCE with unprivileged eBPF and SMT\n");
 
 	return sprintf(buf, "%s%s%s%s%s%s\n",
 		       spectre_v2_strings[spectre_v2_enabled],

@@ -311,5 +311,63 @@ asmlinkage void __arm_smccc_hvc(unsigned
 #define SMCCC_RET_NOT_SUPPORTED			-1
 #define SMCCC_RET_NOT_REQUIRED			-2
 
+/*
+ * Like arm_smccc_1_1* but always returns SMCCC_RET_NOT_SUPPORTED.
+ * Used when the SMCCC conduit is not defined. The empty asm statement
+ * avoids compiler warnings about unused variables.
+ */
+#define __fail_smccc_1_1(...)						\
+	do {								\
+		__declare_args(__count_args(__VA_ARGS__), __VA_ARGS__);	\
+		asm ("" __constraints(__count_args(__VA_ARGS__)));	\
+		if (___res)						\
+			___res->a0 = SMCCC_RET_NOT_SUPPORTED;		\
+	} while (0)
+
+/*
+ * arm_smccc_1_1_invoke() - make an SMCCC v1.1 compliant call
+ *
+ * This is a variadic macro taking one to eight source arguments, and
+ * an optional return structure.
+ *
+ * @a0-a7: arguments passed in registers 0 to 7
+ * @res: result values from registers 0 to 3
+ *
+ * This macro will make either an HVC call or an SMC call depending on the
+ * current SMCCC conduit. If no valid conduit is available then -1
+ * (SMCCC_RET_NOT_SUPPORTED) is returned in @res.a0 (if supplied).
+ *
+ * The return value also provides the conduit that was used.
+ */
+#define arm_smccc_1_1_invoke(...) ({					\
+		int method = arm_smccc_1_1_get_conduit();		\
+		switch (method) {					\
+		case SMCCC_CONDUIT_HVC:					\
+			arm_smccc_1_1_hvc(__VA_ARGS__);			\
+			break;						\
+		case SMCCC_CONDUIT_SMC:					\
+			arm_smccc_1_1_smc(__VA_ARGS__);			\
+			break;						\
+		default:						\
+			__fail_smccc_1_1(__VA_ARGS__);			\
+			method = SMCCC_CONDUIT_NONE;			\
+			break;						\
+		}							\
+		method;							\
+	})
+
+/* Paravirtualised time calls (defined by ARM DEN0057A) */
+#define ARM_SMCCC_HV_PV_TIME_FEATURES				\
+	ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL,			\
+			   ARM_SMCCC_SMC_64,			\
+			   ARM_SMCCC_OWNER_STANDARD_HYP,	\
+			   0x20)
+
+#define ARM_SMCCC_HV_PV_TIME_ST					\
+	ARM_SMCCC_CALL_VAL(ARM_SMCCC_FAST_CALL,			\
+			   ARM_SMCCC_SMC_64,			\
+			   ARM_SMCCC_OWNER_STANDARD_HYP,	\
+			   0x21)
+
 #endif /*__ASSEMBLY__*/
 #endif /*__LINUX_ARM_SMCCC_H*/

@@ -64,6 +64,21 @@ struct psci_operations psci_ops = {
 	.smccc_version = SMCCC_VERSION_1_0,
 };
 
+enum arm_smccc_conduit arm_smccc_1_1_get_conduit(void)
+{
+	if (psci_ops.smccc_version < SMCCC_VERSION_1_1)
+		return SMCCC_CONDUIT_NONE;
+
+	switch (psci_ops.conduit) {
+	case PSCI_CONDUIT_SMC:
+		return SMCCC_CONDUIT_SMC;
+	case PSCI_CONDUIT_HVC:
+		return SMCCC_CONDUIT_HVC;
+	default:
+		return SMCCC_CONDUIT_NONE;
+	}
+}
+
 typedef unsigned long (psci_fn)(unsigned long, unsigned long,
 				unsigned long, unsigned long);
 static psci_fn *invoke_psci_fn;

@@ -89,6 +89,22 @@
 
 #include <linux/linkage.h>
 #include <linux/types.h>
+
+enum arm_smccc_conduit {
+	SMCCC_CONDUIT_NONE,
+	SMCCC_CONDUIT_SMC,
+	SMCCC_CONDUIT_HVC,
+};
+
+/**
+ * arm_smccc_1_1_get_conduit()
+ *
+ * Returns the conduit to be used for SMCCCv1.1 or later.
+ *
+ * When SMCCCv1.1 is not present, returns SMCCC_CONDUIT_NONE.
+ */
+enum arm_smccc_conduit arm_smccc_1_1_get_conduit(void);
+
 /**
  * struct arm_smccc_res - Result from SMC/HVC call
  * @a0-a3 result values from registers 0 to 3

@@ -0,0 +1,28 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+
+#ifndef __ASM_SPECTRE_H
+#define __ASM_SPECTRE_H
+
+enum {
+	SPECTRE_UNAFFECTED,
+	SPECTRE_MITIGATED,
+	SPECTRE_VULNERABLE,
+};
+
+enum {
+	__SPECTRE_V2_METHOD_BPIALL,
+	__SPECTRE_V2_METHOD_ICIALLU,
+	__SPECTRE_V2_METHOD_SMC,
+	__SPECTRE_V2_METHOD_HVC,
+};
+
+enum {
+	SPECTRE_V2_METHOD_BPIALL = BIT(__SPECTRE_V2_METHOD_BPIALL),
+	SPECTRE_V2_METHOD_ICIALLU = BIT(__SPECTRE_V2_METHOD_ICIALLU),
+	SPECTRE_V2_METHOD_SMC = BIT(__SPECTRE_V2_METHOD_SMC),
+	SPECTRE_V2_METHOD_HVC = BIT(__SPECTRE_V2_METHOD_HVC),
+};
+
+void spectre_v2_update_state(unsigned int state, unsigned int methods);
+
+#endif

54 lines not on this page · raw message

203 lines not on this page · raw message

@@ -819,10 +819,22 @@ static inline void __init kuser_init(voi
 }
 #endif
 
+#ifndef CONFIG_CPU_V7M
+static void copy_from_lma(void *vma, void *lma_start, void *lma_end)
+{
+	memcpy(vma, lma_start, lma_end - lma_start);
+}
+
+static void flush_vectors(void *vma, size_t offset, size_t size)
+{
+	unsigned long start = (unsigned long)vma + offset;
+	unsigned long end = start + size;
+
+	flush_icache_range(start, end);
+}
+
 void __init early_trap_init(void *vectors_base)
 {
-#ifndef CONFIG_CPU_V7M
-	unsigned long vectors = (unsigned long)vectors_base;
 	extern char __stubs_start[], __stubs_end[];
 	extern char __vectors_start[], __vectors_end[];
 	unsigned i;

@@ -843,17 +855,20 @@ void __init early_trap_init(void *vector
 	 * into the vector page, mapped at 0xffff0000, and ensure these
 	 * are visible to the instruction stream.
 	 */
-	memcpy((void *)vectors, __vectors_start, __vectors_end - __vectors_start);
-	memcpy((void *)vectors + 0x1000, __stubs_start, __stubs_end - __stubs_start);
+	copy_from_lma(vectors_base, __vectors_start, __vectors_end);
+	copy_from_lma(vectors_base + 0x1000, __stubs_start, __stubs_end);
 
 	kuser_init(vectors_base);
 
-	flush_icache_range(vectors, vectors + PAGE_SIZE * 2);
+	flush_vectors(vectors_base, 0, PAGE_SIZE * 2);
+}
 #else /* ifndef CONFIG_CPU_V7M */
+void __init early_trap_init(void *vectors_base)
+{
 	/*
 	 * on V7-M there is no need to copy the vector table to a dedicated
 	 * memory area. The address is configurable and so a table in the kernel
 	 * image can be used.
 	 */
-#endif
 }
+#endif

@@ -108,6 +108,16 @@
 	.endm
 #endif
 
+#if __LINUX_ARM_ARCH__ < 7
+	.macro	dsb, args
+	mcr	p15, 0, r0, c7, c10, 4
+	.endm
+
+	.macro	isb, args
+	mcr	p15, 0, r0, c7, r5, 4
+	.endm
+#endif
+
 	.macro asm_trace_hardirqs_off, save=1
 #if defined(CONFIG_TRACE_IRQFLAGS)
 	.if \save

@@ -14,6 +14,7 @@ enum {
 	__SPECTRE_V2_METHOD_ICIALLU,
 	__SPECTRE_V2_METHOD_SMC,
 	__SPECTRE_V2_METHOD_HVC,
+	__SPECTRE_V2_METHOD_LOOP8,
 };
 
 enum {

@@ -21,8 +22,11 @@ enum {
 	SPECTRE_V2_METHOD_ICIALLU = BIT(__SPECTRE_V2_METHOD_ICIALLU),
 	SPECTRE_V2_METHOD_SMC = BIT(__SPECTRE_V2_METHOD_SMC),
 	SPECTRE_V2_METHOD_HVC = BIT(__SPECTRE_V2_METHOD_HVC),
+	SPECTRE_V2_METHOD_LOOP8 = BIT(__SPECTRE_V2_METHOD_LOOP8),
 };
 
 void spectre_v2_update_state(unsigned int state, unsigned int methods);
 
+int spectre_bhb_update_vectors(unsigned int method);
+
 #endif

@@ -45,6 +45,10 @@ ssize_t cpu_show_spectre_v2(struct devic
 		method = "Firmware call";
 		break;
 
+	case SPECTRE_V2_METHOD_LOOP8:
+		method = "History overwrite";
+		break;
+
 	default:
 		method = "Multiple mitigations";
 		break;

@@ -31,6 +31,7 @@
 #include <linux/atomic.h>
 #include <asm/cacheflush.h>
 #include <asm/exception.h>
+#include <asm/spectre.h>
 #include <asm/unistd.h>
 #include <asm/traps.h>
 #include <asm/ptrace.h>

@@ -833,6 +834,43 @@ static void flush_vectors(void *vma, siz
 	flush_icache_range(start, end);
 }
 
+#ifdef CONFIG_HARDEN_BRANCH_HISTORY
+int spectre_bhb_update_vectors(unsigned int method)
+{
+	extern char __vectors_bhb_bpiall_start[], __vectors_bhb_bpiall_end[];
+	extern char __vectors_bhb_loop8_start[], __vectors_bhb_loop8_end[];
+	void *vec_start, *vec_end;
+
+	if (system_state >= SYSTEM_RUNNING) {
+		pr_err("CPU%u: Spectre BHB workaround too late - system vulnerable\n",
+		       smp_processor_id());
+		return SPECTRE_VULNERABLE;
+	}
+
+	switch (method) {
+	case SPECTRE_V2_METHOD_LOOP8:
+		vec_start = __vectors_bhb_loop8_start;
+		vec_end = __vectors_bhb_loop8_end;
+		break;
+
+	case SPECTRE_V2_METHOD_BPIALL:
+		vec_start = __vectors_bhb_bpiall_start;
+		vec_end = __vectors_bhb_bpiall_end;
+		break;
+
+	default:
+		pr_err("CPU%u: unknown Spectre BHB state %d\n",
+		       smp_processor_id(), method);
+		return SPECTRE_VULNERABLE;
+	}
+
+	copy_from_lma(vectors_page, vec_start, vec_end);
+	flush_vectors(vectors_page, 0, vec_end - vec_start);
+
+	return SPECTRE_MITIGATED;
+}
+#endif
+
 void __init early_trap_init(void *vectors_base)
 {
 	extern char __stubs_start[], __stubs_end[];

86 lines not on this page · raw message

@@ -1,9 +1,19 @@
 // SPDX-License-Identifier: GPL-2.0-only
+#include <linux/bpf.h>
 #include <linux/cpu.h>
 #include <linux/device.h>
 
 #include <asm/spectre.h>
 
+static bool _unprivileged_ebpf_enabled(void)
+{
+#ifdef CONFIG_BPF_SYSCALL
+	return !sysctl_unprivileged_bpf_disabled;
+#else
+	return false
+#endif
+}
+
 ssize_t cpu_show_spectre_v1(struct device *dev, struct device_attribute *attr,
 			    char *buf)
 {

@@ -31,6 +41,9 @@ ssize_t cpu_show_spectre_v2(struct devic
 	if (spectre_v2_state != SPECTRE_MITIGATED)
 		return sprintf(buf, "%s\n", "Vulnerable");
 
+	if (_unprivileged_ebpf_enabled())
+		return sprintf(buf, "Vulnerable: Unprivileged eBPF enabled\n");
+
 	switch (spectre_v2_methods) {
 	case SPECTRE_V2_METHOD_BPIALL:
 		method = "Branch predictor hardening";

@@ -10,7 +10,7 @@ static bool _unprivileged_ebpf_enabled(v
 #ifdef CONFIG_BPF_SYSCALL
 	return !sysctl_unprivileged_bpf_disabled;
 #else
-	return false
+	return false;
 #endif
 }
 

@@ -114,7 +114,7 @@
 	.endm
 
 	.macro	isb, args
-	mcr	p15, 0, r0, c7, r5, 4
+	mcr	p15, 0, r0, c7, c5, 4
 	.endm
 #endif
 

@@ -586,7 +586,7 @@ static enum spectre_v2_user_mitigation s
 static enum spectre_v2_user_mitigation spectre_v2_user_ibpb __ro_after_init =
 	SPECTRE_V2_USER_NONE;
 
-#ifdef RETPOLINE
+#ifdef CONFIG_RETPOLINE
 static bool spectre_v2_bad_module;
 
 bool retpoline_module_ok(bool has_retpoline)

@@ -107,7 +107,7 @@
 #define __weak		__attribute__((weak))
 #define __alias(symbol)	__attribute__((alias(#symbol)))
 
-#ifdef RETPOLINE
+#ifdef CONFIG_RETPOLINE
 #define __noretpoline __attribute__((indirect_branch("keep")))
 #endif
 

@@ -791,7 +791,7 @@ static inline void module_bug_finalize(c
 static inline void module_bug_cleanup(struct module *mod) {}
 #endif	/* CONFIG_GENERIC_BUG */
 
-#ifdef RETPOLINE
+#ifdef CONFIG_RETPOLINE
 extern bool retpoline_module_ok(bool has_retpoline);
 #else
 static inline bool retpoline_module_ok(bool has_retpoline)

@@ -2147,7 +2147,7 @@ static void add_intree_flag(struct buffe
 /* Cannot check for assembler */
 static void add_retpoline(struct buffer *b)
 {
-	buf_printf(b, "\n#ifdef RETPOLINE\n");
+	buf_printf(b, "\n#ifdef CONFIG_RETPOLINE\n");
 	buf_printf(b, "MODULE_INFO(retpoline, \"Y\");\n");
 	buf_printf(b, "#endif\n");
 }

@@ -110,7 +110,8 @@ static unsigned int spectre_v2_install_w
 #else
 static unsigned int spectre_v2_install_workaround(unsigned int method)
 {
-	pr_info("CPU%u: Spectre V2: workarounds disabled by configuration\n");
+	pr_info("CPU%u: Spectre V2: workarounds disabled by configuration\n",
+		smp_processor_id());
 
 	return SPECTRE_VULNERABLE;
 }

@@ -387,7 +387,14 @@ int xenbus_grant_ring(struct xenbus_devi
 		      unsigned int nr_pages, grant_ref_t *grefs)
 {
 	int err;
-	int i, j;
+	unsigned int i;
+	grant_ref_t gref_head;
+
+	err = gnttab_alloc_grant_references(nr_pages, &gref_head);
+	if (err) {
+		xenbus_dev_fatal(dev, err, "granting access to ring page");
+		return err;
+	}
 
 	for (i = 0; i < nr_pages; i++) {
 		unsigned long gfn;

@@ -397,23 +404,14 @@ int xenbus_grant_ring(struct xenbus_devi
 		else
 			gfn = virt_to_gfn(vaddr);
 
-		err = gnttab_grant_foreign_access(dev->otherend_id, gfn, 0);
-		if (err < 0) {
-			xenbus_dev_fatal(dev, err,
-					 "granting access to ring page");
-			goto fail;
-		}
-		grefs[i] = err;
+		grefs[i] = gnttab_claim_grant_reference(&gref_head);
+		gnttab_grant_foreign_access_ref(grefs[i], dev->otherend_id,
+						gfn, 0);
 
 		vaddr = vaddr + XEN_PAGE_SIZE;
 	}
 
 	return 0;
-
-fail:
-	for (j = 0; j < i; j++)
-		gnttab_end_foreign_access_ref(grefs[j], 0);
-	return err;
 }
 EXPORT_SYMBOL_GPL(xenbus_grant_ring);
 

@@ -377,11 +377,21 @@ static void gnttab_add_deferred(grant_re
 	       what, ref, page ? page_to_pfn(page) : -1);
 }
 
+int gnttab_try_end_foreign_access(grant_ref_t ref)
+{
+	int ret = _gnttab_end_foreign_access_ref(ref, 0);
+
+	if (ret)
+		put_free_entry(ref);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(gnttab_try_end_foreign_access);
+
 void gnttab_end_foreign_access(grant_ref_t ref, int readonly,
 			       unsigned long page)
 {
-	if (gnttab_end_foreign_access_ref(ref, readonly)) {
-		put_free_entry(ref);
+	if (gnttab_try_end_foreign_access(ref)) {
 		if (page != 0)
 			put_page(virt_to_page(page));
 	} else

@@ -97,10 +97,22 @@ int gnttab_end_foreign_access_ref(grant_
  * access has been ended, free the given page too.  Access will be ended
  * immediately iff the grant entry is not in use, otherwise it will happen
  * some time later.  page may be 0, in which case no freeing will occur.
+ * Note that the granted page might still be accessed (read or write) by the
+ * other side after gnttab_end_foreign_access() returns, so even if page was
+ * specified as 0 it is not allowed to just reuse the page for other
+ * purposes immediately.
  */
 void gnttab_end_foreign_access(grant_ref_t ref, int readonly,
 			       unsigned long page);
 
+/*
+ * End access through the given grant reference, iff the grant entry is
+ * no longer in use.  In case of success ending foreign access, the
+ * grant reference is deallocated.
+ * Return 1 if the grant entry was freed, 0 if it is still in use.
+ */
+int gnttab_try_end_foreign_access(grant_ref_t ref);
+
 int gnttab_grant_foreign_transfer(domid_t domid, unsigned long pfn);
 
 unsigned long gnttab_end_foreign_transfer_ref(grant_ref_t ref);

@@ -1266,17 +1266,16 @@ static void blkif_free_ring(struct blkfr
 		list_for_each_entry_safe(persistent_gnt, n,
 					 &rinfo->grants, node) {
 			list_del(&persistent_gnt->node);
-			if (persistent_gnt->gref != GRANT_INVALID_REF) {
-				gnttab_end_foreign_access(persistent_gnt->gref,
-							  0, 0UL);
-				rinfo->persistent_gnts_c--;
-			}
+			if (persistent_gnt->gref == GRANT_INVALID_REF ||
+			    !gnttab_try_end_foreign_access(persistent_gnt->gref))
+				continue;
+
+			rinfo->persistent_gnts_c--;
 			if (info->feature_persistent)
 				__free_page(persistent_gnt->page);
 			kfree(persistent_gnt);
 		}
 	}
-	BUG_ON(rinfo->persistent_gnts_c != 0);
 
 	for (i = 0; i < BLK_RING_SIZE(info); i++) {
 		/*

@@ -1333,7 +1332,8 @@ free_shadow:
 			rinfo->ring_ref[i] = GRANT_INVALID_REF;
 		}
 	}
-	free_pages((unsigned long)rinfo->ring.sring, get_order(info->nr_ring_pages * XEN_PAGE_SIZE));
+	free_pages_exact(rinfo->ring.sring,
+			 info->nr_ring_pages * XEN_PAGE_SIZE);
 	rinfo->ring.sring = NULL;
 
 	if (rinfo->irq)

@@ -1417,9 +1417,15 @@ static int blkif_get_final_status(enum b
 	return BLKIF_RSP_OKAY;
 }
 
-static bool blkif_completion(unsigned long *id,
-			     struct blkfront_ring_info *rinfo,
-			     struct blkif_response *bret)
+/*
+ * Return values:
+ *  1 response processed.
+ *  0 missing further responses.
+ * -1 error while processing.
+ */
+static int blkif_completion(unsigned long *id,
+			    struct blkfront_ring_info *rinfo,
+			    struct blkif_response *bret)
 {
 	int i = 0;
 	struct scatterlist *sg;

@@ -1493,42 +1499,43 @@ static bool blkif_completion(unsigned lo
 	}
 	/* Add the persistent grant into the list of free grants */
 	for (i = 0; i < num_grant; i++) {
-		if (gnttab_query_foreign_access(s->grants_used[i]->gref)) {
+		if (!gnttab_try_end_foreign_access(s->grants_used[i]->gref)) {
 			/*
 			 * If the grant is still mapped by the backend (the
 			 * backend has chosen to make this grant persistent)
 			 * we add it at the head of the list, so it will be
 			 * reused first.
 			 */
-			if (!info->feature_persistent)
-				pr_alert_ratelimited("backed has not unmapped grant: %u\n",
-						     s->grants_used[i]->gref);
+			if (!info->feature_persistent) {
+				pr_alert("backed has not unmapped grant: %u\n",
+					 s->grants_used[i]->gref);
+				return -1;
+			}
 			list_add(&s->grants_used[i]->node, &rinfo->grants);
 			rinfo->persistent_gnts_c++;
 		} else {
 			/*
-			 * If the grant is not mapped by the backend we end the
-			 * foreign access and add it to the tail of the list,
-			 * so it will not be picked again unless we run out of
-			 * persistent grants.
+			 * If the grant is not mapped by the backend we add it
+			 * to the tail of the list, so it will not be picked
+			 * again unless we run out of persistent grants.
 			 */
-			gnttab_end_foreign_access(s->grants_used[i]->gref, 0, 0UL);
 			s->grants_used[i]->gref = GRANT_INVALID_REF;
 			list_add_tail(&s->grants_used[i]->node, &rinfo->grants);
 		}
 	}
 	if (s->req.operation == BLKIF_OP_INDIRECT) {
 		for (i = 0; i < INDIRECT_GREFS(num_grant); i++) {
-			if (gnttab_query_foreign_access(s->indirect_grants[i]->gref)) {
-				if (!info->feature_persistent)
-					pr_alert_ratelimited("backed has not unmapped grant: %u\n",
-							     s->indirect_grants[i]->gref);
+			if (!gnttab_try_end_foreign_access(s->indirect_grants[i]->gref)) {
+				if (!info->feature_persistent) {
+					pr_alert("backed has not unmapped grant: %u\n",
+						 s->indirect_grants[i]->gref);
+					return -1;
+				}
 				list_add(&s->indirect_grants[i]->node, &rinfo->grants);
 				rinfo->persistent_gnts_c++;
 			} else {
 				struct page *indirect_page;
 
-				gnttab_end_foreign_access(s->indirect_grants[i]->gref, 0, 0UL);
 				/*
 				 * Add the used indirect page back to the list of
 				 * available pages for indirect grefs.

@@ -1610,12 +1617,17 @@ static irqreturn_t blkif_interrupt(int i
 		}
 
 		if (bret.operation != BLKIF_OP_DISCARD) {
+			int ret;
+
 			/*
 			 * We may need to wait for an extra response if the
 			 * I/O request is split in 2
 			 */
-			if (!blkif_completion(&id, rinfo, &bret))
+			ret = blkif_completion(&id, rinfo, &bret);
+			if (!ret)
 				continue;
+			if (unlikely(ret < 0))
+				goto err;
 		}
 
 		if (add_id_to_freelist(rinfo, id)) {

@@ -1717,8 +1729,7 @@ static int setup_blkring(struct xenbus_d
 	for (i = 0; i < info->nr_ring_pages; i++)
 		rinfo->ring_ref[i] = GRANT_INVALID_REF;
 
-	sring = (struct blkif_sring *)__get_free_pages(GFP_NOIO | __GFP_HIGH,
-						       get_order(ring_size));
+	sring = alloc_pages_exact(ring_size, GFP_NOIO);
 	if (!sring) {
 		xenbus_dev_fatal(dev, -ENOMEM, "allocating shared ring");
 		return -ENOMEM;

@@ -1728,7 +1739,7 @@ static int setup_blkring(struct xenbus_d
 
 	err = xenbus_grant_ring(dev, rinfo->ring.sring, info->nr_ring_pages, gref);
 	if (err < 0) {
-		free_pages((unsigned long)sring, get_order(ring_size));
+		free_pages_exact(sring, ring_size);
 		rinfo->ring.sring = NULL;
 		goto fail;
 	}

@@ -413,14 +413,12 @@ static bool xennet_tx_buf_gc(struct netf
 			queue->tx_link[id] = TX_LINK_NONE;
 			skb = queue->tx_skbs[id];
 			queue->tx_skbs[id] = NULL;
-			if (unlikely(gnttab_query_foreign_access(
-				queue->grant_tx_ref[id]) != 0)) {
+			if (unlikely(!gnttab_end_foreign_access_ref(
+				queue->grant_tx_ref[id], GNTMAP_readonly))) {
 				dev_alert(dev,
 					  "Grant still in use by backend domain\n");
 				goto err;
 			}
-			gnttab_end_foreign_access_ref(
-				queue->grant_tx_ref[id], GNTMAP_readonly);
 			gnttab_release_grant_reference(
 				&queue->gref_tx_head, queue->grant_tx_ref[id]);
 			queue->grant_tx_ref[id] = GRANT_INVALID_REF;

@@ -210,12 +210,11 @@ static void scsifront_gnttab_done(struct
 		return;
 
 	for (i = 0; i < s->nr_grants; i++) {
-		if (unlikely(gnttab_query_foreign_access(s->gref[i]) != 0)) {
+		if (unlikely(!gnttab_try_end_foreign_access(s->gref[i]))) {
 			shost_printk(KERN_ALERT, info->host, KBUILD_MODNAME
 				     "grant still in use by backend\n");
 			BUG();
 		}
-		gnttab_end_foreign_access(s->gref[i], 0, 0UL);
 	}
 
 	kfree(s->sg);

@@ -166,20 +166,14 @@ undo:
 		__del_gref(gref);
 	}
 
-	/* It's possible for the target domain to map the just-allocated grant
-	 * references by blindly guessing their IDs; if this is done, then
-	 * __del_gref will leave them in the queue_gref list. They need to be
-	 * added to the global list so that we can free them when they are no
-	 * longer referenced.
-	 */
-	if (unlikely(!list_empty(&queue_gref)))
-		list_splice_tail(&queue_gref, &gref_list);
 	mutex_unlock(&gref_mutex);
 	return rc;
 }
 
 static void __del_gref(struct gntalloc_gref *gref)
 {
+	unsigned long addr;
+
 	if (gref->notify.flags & UNMAP_NOTIFY_CLEAR_BYTE) {
 		uint8_t *tmp = kmap(gref->page);
 		tmp[gref->notify.pgoff] = 0;

@@ -193,21 +187,16 @@ static void __del_gref(struct gntalloc_g
 	gref->notify.flags = 0;
 
 	if (gref->gref_id) {
-		if (gnttab_query_foreign_access(gref->gref_id))
-			return;
-
-		if (!gnttab_end_foreign_access_ref(gref->gref_id, 0))
-			return;
-
-		gnttab_free_grant_reference(gref->gref_id);
+		if (gref->page) {
+			addr = (unsigned long)page_to_virt(gref->page);
+			gnttab_end_foreign_access(gref->gref_id, 0, addr);
+		} else
+			gnttab_free_grant_reference(gref->gref_id);
 	}
 
 	gref_size--;
 	list_del(&gref->next_gref);
 
-	if (gref->page)
-		__free_page(gref->page);
-
 	kfree(gref);
 }
 

@@ -113,13 +113,6 @@ struct gnttab_ops {
 	 * return the frame.
 	 */
 	unsigned long (*end_foreign_transfer_ref)(grant_ref_t ref);
-	/*
-	 * Query the status of a grant entry. Ref parameter is reference of
-	 * queried grant entry, return value is the status of queried entry.
-	 * Detailed status(writing/reading) can be gotten from the return value
-	 * by bit operations.
-	 */
-	int (*query_foreign_access)(grant_ref_t ref);
 };
 
 struct unmap_refs_callback_data {

@@ -254,17 +247,6 @@ int gnttab_grant_foreign_access(domid_t
 }
 EXPORT_SYMBOL_GPL(gnttab_grant_foreign_access);
 
-static int gnttab_query_foreign_access_v1(grant_ref_t ref)
-{
-	return gnttab_shared.v1[ref].flags & (GTF_reading|GTF_writing);
-}
-
-int gnttab_query_foreign_access(grant_ref_t ref)
-{
-	return gnttab_interface->query_foreign_access(ref);
-}
-EXPORT_SYMBOL_GPL(gnttab_query_foreign_access);
-
 static int gnttab_end_foreign_access_ref_v1(grant_ref_t ref, int readonly)
 {
 	u16 flags, nflags;

@@ -1028,7 +1010,6 @@ static const struct gnttab_ops gnttab_v1
 	.update_entry			= gnttab_update_entry_v1,
 	.end_foreign_access_ref		= gnttab_end_foreign_access_ref_v1,
 	.end_foreign_transfer_ref	= gnttab_end_foreign_transfer_ref_v1,
-	.query_foreign_access		= gnttab_query_foreign_access_v1,
 };
 
 static void gnttab_request_version(void)

@@ -118,8 +118,6 @@ int gnttab_grant_foreign_transfer(domid_
 unsigned long gnttab_end_foreign_transfer_ref(grant_ref_t ref);
 unsigned long gnttab_end_foreign_transfer(grant_ref_t ref);
 
-int gnttab_query_foreign_access(grant_ref_t ref);
-
 /*
  * operations on reserved batches of grant references
  */

@@ -113,6 +113,10 @@ struct gnttab_ops {
 	 * return the frame.
 	 */
 	unsigned long (*end_foreign_transfer_ref)(grant_ref_t ref);
+	/*
+	 * Read the frame number related to a given grant reference.
+	 */
+	unsigned long (*read_frame)(grant_ref_t ref);
 };
 
 struct unmap_refs_callback_data {

@@ -277,6 +281,11 @@ int gnttab_end_foreign_access_ref(grant_
 }
 EXPORT_SYMBOL_GPL(gnttab_end_foreign_access_ref);
 
+static unsigned long gnttab_read_frame_v1(grant_ref_t ref)
+{
+	return gnttab_shared.v1[ref].frame;
+}
+
 struct deferred_entry {
 	struct list_head list;
 	grant_ref_t ref;

@@ -306,12 +315,9 @@ static void gnttab_handle_deferred(unsig
 		spin_unlock_irqrestore(&gnttab_list_lock, flags);
 		if (_gnttab_end_foreign_access_ref(entry->ref, entry->ro)) {
 			put_free_entry(entry->ref);
-			if (entry->page) {
-				pr_debug("freeing g.e. %#x (pfn %#lx)\n",
-					 entry->ref, page_to_pfn(entry->page));
-				put_page(entry->page);
-			} else
-				pr_info("freeing g.e. %#x\n", entry->ref);
+			pr_debug("freeing g.e. %#x (pfn %#lx)\n",
+				 entry->ref, page_to_pfn(entry->page));
+			put_page(entry->page);
 			kfree(entry);
 			entry = NULL;
 		} else {

@@ -336,9 +342,18 @@ static void gnttab_handle_deferred(unsig
 static void gnttab_add_deferred(grant_ref_t ref, bool readonly,
 				struct page *page)
 {
-	struct deferred_entry *entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
+	struct deferred_entry *entry;
+	gfp_t gfp = (in_atomic() || irqs_disabled()) ? GFP_ATOMIC : GFP_KERNEL;
 	const char *what = KERN_WARNING "leaking";
 
+	entry = kmalloc(sizeof(*entry), gfp);
+	if (!page) {
+		unsigned long gfn = gnttab_interface->read_frame(ref);
+
+		page = pfn_to_page(gfn_to_pfn(gfn));
+		get_page(page);
+	}
+
 	if (entry) {
 		unsigned long flags;
 

@@ -1010,6 +1025,7 @@ static const struct gnttab_ops gnttab_v1
 	.update_entry			= gnttab_update_entry_v1,
 	.end_foreign_access_ref		= gnttab_end_foreign_access_ref_v1,
 	.end_foreign_transfer_ref	= gnttab_end_foreign_transfer_ref_v1,
+	.read_frame			= gnttab_read_frame_v1,
 };
 
 static void gnttab_request_version(void)

@@ -100,7 +100,12 @@ int gnttab_end_foreign_access_ref(grant_
  * Note that the granted page might still be accessed (read or write) by the
  * other side after gnttab_end_foreign_access() returns, so even if page was
  * specified as 0 it is not allowed to just reuse the page for other
- * purposes immediately.
+ * purposes immediately. gnttab_end_foreign_access() will take an additional
+ * reference to the granted page in this case, which is dropped only after
+ * the grant is no longer in use.
+ * This requires that multi page allocations for areas subject to
+ * gnttab_end_foreign_access() are done via alloc_pages_exact() (and freeing
+ * via free_pages_exact()) in order to avoid high order pages.
  */
 void gnttab_end_foreign_access(grant_ref_t ref, int readonly,
 			       unsigned long page);

@@ -838,7 +838,6 @@ static int xennet_get_responses(struct n
 	int max = XEN_NETIF_NR_SLOTS_MIN + (rx->status <= RX_COPY_THRESHOLD);
 	int slots = 1;
 	int err = 0;
-	unsigned long ret;
 
 	if (rx->flags & XEN_NETRXF_extra_info) {
 		err = xennet_get_extras(queue, extras, rp);

@@ -869,8 +868,13 @@ static int xennet_get_responses(struct n
 			goto next;
 		}
 
-		ret = gnttab_end_foreign_access_ref(ref, 0);
-		BUG_ON(!ret);
+		if (!gnttab_end_foreign_access_ref(ref, 0)) {
+			dev_alert(dev,
+				  "Grant still in use by backend domain\n");
+			queue->info->broken = true;
+			dev_alert(dev, "Disabled for further use\n");
+			return -EINVAL;
+		}
 
 		gnttab_release_grant_reference(&queue->gref_rx_head, ref);
 

@@ -1074,6 +1078,10 @@ static int xennet_poll(struct napi_struc
 		err = xennet_get_responses(queue, &rinfo, rp, &tmpq);
 
 		if (unlikely(err)) {
+			if (queue->info->broken) {
+				spin_unlock(&queue->rx_lock);
+				return 0;
+			}
 err:
 			while ((skb = __skb_dequeue(&tmpq)))
 				__skb_queue_tail(&errq, skb);

@@ -1671,7 +1679,7 @@ static int setup_netfront(struct xenbus_
 			struct netfront_queue *queue, unsigned int feature_split_evtchn)
 {
 	struct xen_netif_tx_sring *txs;
-	struct xen_netif_rx_sring *rxs;
+	struct xen_netif_rx_sring *rxs = NULL;
 	grant_ref_t gref;
 	int err;
 

@@ -1691,21 +1699,21 @@ static int setup_netfront(struct xenbus_
 
 	err = xenbus_grant_ring(dev, txs, 1, &gref);
 	if (err < 0)
-		goto grant_tx_ring_fail;
+		goto fail;
 	queue->tx_ring_ref = gref;
 
 	rxs = (struct xen_netif_rx_sring *)get_zeroed_page(GFP_NOIO | __GFP_HIGH);
 	if (!rxs) {
 		err = -ENOMEM;
 		xenbus_dev_fatal(dev, err, "allocating rx ring page");
-		goto alloc_rx_ring_fail;
+		goto fail;
 	}
 	SHARED_RING_INIT(rxs);
 	FRONT_RING_INIT(&queue->rx, rxs, XEN_PAGE_SIZE);
 
 	err = xenbus_grant_ring(dev, rxs, 1, &gref);
 	if (err < 0)
-		goto grant_rx_ring_fail;
+		goto fail;
 	queue->rx_ring_ref = gref;
 
 	if (feature_split_evtchn)

@@ -1718,22 +1726,28 @@ static int setup_netfront(struct xenbus_
 		err = setup_netfront_single(queue);
 
 	if (err)
-		goto alloc_evtchn_fail;
+		goto fail;
 
 	return 0;
 
 	/* If we fail to setup netfront, it is safe to just revoke access to
 	 * granted pages because backend is not accessing it at this point.
 	 */
-alloc_evtchn_fail:
-	gnttab_end_foreign_access_ref(queue->rx_ring_ref, 0);
-grant_rx_ring_fail:
-	free_page((unsigned long)rxs);
-alloc_rx_ring_fail:
-	gnttab_end_foreign_access_ref(queue->tx_ring_ref, 0);
-grant_tx_ring_fail:
-	free_page((unsigned long)txs);
-fail:
+ fail:
+	if (queue->rx_ring_ref != GRANT_INVALID_REF) {
+		gnttab_end_foreign_access(queue->rx_ring_ref, 0,
+					  (unsigned long)rxs);
+		queue->rx_ring_ref = GRANT_INVALID_REF;
+	} else {
+		free_page((unsigned long)rxs);
+	}
+	if (queue->tx_ring_ref != GRANT_INVALID_REF) {
+		gnttab_end_foreign_access(queue->tx_ring_ref, 0,
+					  (unsigned long)txs);
+		queue->tx_ring_ref = GRANT_INVALID_REF;
+	} else {
+		free_page((unsigned long)txs);
+	}
 	return err;
 }