[PULL 5/5] sgx: Reset the vEPC regions during VM reboot

[Paolo Bonzini <pbonzini@redhat.com>]

From: Yang Zhong <yang.zhong@intel.com>

For bare-metal SGX on real hardware, the hardware provides guarantees
SGX state at reboot.  For instance, all pages start out uninitialized.
The vepc driver provides a similar guarantee today for freshly-opened
vepc instances, but guests such as Windows expect all pages to be in
uninitialized state on startup, including after every guest reboot.

Qemu can invoke the ioctl to bring its vEPC pages back to uninitialized
state. There is a possibility that some pages fail to be removed if they
are SECS pages, and the child and SECS pages could be in separate vEPC
regions.  Therefore, the ioctl returns the number of EREMOVE failures,
telling Qemu to try the ioctl again after it's done with all vEPC regions.

The related kernel patches:
Link: https://lkml.kernel.org/r/20211021201155.1523989-3-pbonzini@redhat.com

Signed-off-by: Yang Zhong <yang.zhong@intel.com>
Message-Id: <20211101162009.62161-6-yang.zhong@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 hw/i386/sgx.c | 50 ++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 50 insertions(+)

diff --git a/hw/i386/sgx.c b/hw/i386/sgx.c
index 11607568b6..8fef3dd8fa 100644
--- a/hw/i386/sgx.c
+++ b/hw/i386/sgx.c
@@ -21,6 +21,8 @@
 #include "qapi/qapi-commands-misc-target.h"
 #include "exec/address-spaces.h"
 #include "sysemu/hw_accel.h"
+#include "sysemu/reset.h"
+#include <sys/ioctl.h>
 
 #define SGX_MAX_EPC_SECTIONS            8
 #define SGX_CPUID_EPC_INVALID           0x0
@@ -29,6 +31,11 @@
 #define SGX_CPUID_EPC_SECTION           0x1
 #define SGX_CPUID_EPC_MASK              0xF
 
+#define SGX_MAGIC 0xA4
+#define SGX_IOC_VEPC_REMOVE_ALL       _IO(SGX_MAGIC, 0x04)
+
+#define RETRY_NUM                       2
+
 static uint64_t sgx_calc_section_metric(uint64_t low, uint64_t high)
 {
     return (low & MAKE_64BIT_MASK(12, 20)) +
@@ -59,6 +66,46 @@ static uint64_t sgx_calc_host_epc_section_size(void)
     return size;
 }
 
+static void sgx_epc_reset(void *opaque)
+{
+    PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
+    HostMemoryBackend *hostmem;
+    SGXEPCDevice *epc;
+    int failures;
+    int fd, i, j, r;
+    static bool warned = false;
+
+    /*
+     * The second pass is needed to remove SECS pages that could not
+     * be removed during the first.
+     */
+    for (i = 0; i < RETRY_NUM; i++) {
+        failures = 0;
+        for (j = 0; j < pcms->sgx_epc.nr_sections; j++) {
+            epc = pcms->sgx_epc.sections[j];
+            hostmem = MEMORY_BACKEND(epc->hostmem);
+            fd = memory_region_get_fd(host_memory_backend_get_memory(hostmem));
+
+            r = ioctl(fd, SGX_IOC_VEPC_REMOVE_ALL);
+            if (r == -ENOTTY && !warned) {
+                warned = true;
+                warn_report("kernel does not support SGX_IOC_VEPC_REMOVE_ALL");
+                warn_report("SGX might operate incorrectly in the guest after reset");
+                break;
+            } else if (r > 0) {
+                /* SECS pages remain */
+                failures++;
+                if (i == 1) {
+                    error_report("cannot reset vEPC section %d", j);
+                }
+            }
+        }
+        if (!failures) {
+            break;
+        }
+     }
+}
+
 SGXInfo *qmp_query_sgx_capabilities(Error **errp)
 {
     SGXInfo *info = NULL;
@@ -190,4 +237,7 @@ void pc_machine_init_sgx_epc(PCMachineState *pcms)
     }
 
     memory_region_set_size(&sgx_epc->mr, sgx_epc->size);
+
+    /* register the reset callback for sgx epc */
+    qemu_register_reset(sgx_epc_reset, NULL);
 }
-- 
2.33.1