]> git.kernelconcepts.de Git - karo-tx-linux.git/blobdiff - arch/x86/kernel/vm86_32.c
Merge remote-tracking branches 'asoc/topic/wm8904', 'asoc/topic/wm8955' and 'asoc...
[karo-tx-linux.git] / arch / x86 / kernel / vm86_32.c
index abd8b856bd2b50af307caa2eaad953dc80c4ed00..5246193519614dbd8d3a602544e8117376df05f7 100644 (file)
@@ -45,6 +45,7 @@
 #include <linux/audit.h>
 #include <linux/stddef.h>
 #include <linux/slab.h>
+#include <linux/security.h>
 
 #include <asm/uaccess.h>
 #include <asm/io.h>
@@ -232,6 +233,32 @@ static long do_sys_vm86(struct vm86plus_struct __user *user_vm86, bool plus)
        struct pt_regs *regs = current_pt_regs();
        unsigned long err = 0;
 
+       err = security_mmap_addr(0);
+       if (err) {
+               /*
+                * vm86 cannot virtualize the address space, so vm86 users
+                * need to manage the low 1MB themselves using mmap.  Given
+                * that BIOS places important data in the first page, vm86
+                * is essentially useless if mmap_min_addr != 0.  DOSEMU,
+                * for example, won't even bother trying to use vm86 if it
+                * can't map a page at virtual address 0.
+                *
+                * To reduce the available kernel attack surface, simply
+                * disallow vm86(old) for users who cannot mmap at va 0.
+                *
+                * The implementation of security_mmap_addr will allow
+                * suitably privileged users to map va 0 even if
+                * vm.mmap_min_addr is set above 0, and we want this
+                * behavior for vm86 as well, as it ensures that legacy
+                * tools like vbetool will not fail just because of
+                * vm.mmap_min_addr.
+                */
+               pr_info_once("Denied a call to vm86(old) from %s[%d] (uid: %d).  Set the vm.mmap_min_addr sysctl to 0 and/or adjust LSM mmap_min_addr policy to enable vm86 if you are using a vm86-based DOS emulator.\n",
+                            current->comm, task_pid_nr(current),
+                            from_kuid_munged(&init_user_ns, current_uid()));
+               return -EPERM;
+       }
+
        if (!vm86) {
                if (!(vm86 = kzalloc(sizeof(*vm86), GFP_KERNEL)))
                        return -ENOMEM;