linux-kernel - Slab-of-out-bounds in search_memslots() in kvm

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <CANhq1J6AJvkXUVZtbYgZubepU8xL88Q56UrcDprmG_eDapmXtA@mail.gmail.com>
Date:   Fri, 17 Apr 2020 21:24:10 +0800
From:   sam hao <ssesamhao@...il.com>
To:     kvm@...r.kernel.org, linux-kernel@...r.kernel.org,
        pbonzini@...hat.com
Subject: Slab-of-out-bounds in search_memslots() in kvm_host.h

Hi,

I've found possible out of bounds access in search_memslots() in kvm_host.h.
In search_memslots(struct kvm_memslots *slots, gfn_t gfn),  a binary
search is used  for slot searching,  as following code shows
        while (start < end) {
                slot = start + (end - start) / 2;

                if (gfn >= memslots[slot].base_gfn)
                        end = slot;
                else
                        start = slot + 1;
        }

        if (gfn >= memslots[start].base_gfn &&
            gfn < memslots[start].base_gfn + memslots[start].npages) {
                atomic_set(&slots->lru_slot, start);
                return &memslots[start];
        }

However, start may equal to slots->used_slots  when gfn is smaller
than every base_gfn, which cause out of bound access in  if condition.
Following code can trigger this bug:

#include <stdint.h>
#include <unistd.h>
#include <linux/kvm.h>
#include <asm/kvm.h>
#include <sys/ioctl.h>
#include <fcntl.h>

int main(int argc, char **agrv){

        struct kvm_userspace_memory_region kvm_userspace_memory_region_0 = {
                .slot = 4098152658,
                .flags = 1653871800,
                .guest_phys_addr = 9228163640593578308,
                .memory_size = 13154652985641659684,
                .userspace_addr = 2934507574655831761
        };
        char *s_0 = "/dev/kvm";
        struct kvm_vapic_addr kvm_vapic_addr_1 = {
                .vapic_addr=4096
        };

        int32_t r0 = open(s_0,0,0);
        int32_t r1 = ioctl(r0,44545,0);
        ioctl(r1,44640);
        ioctl(r1,1075883590,&kvm_userspace_memory_region_0);
        int32_t r2 = ioctl(r1,44609,0);
        ioctl(r2,44672,0);
        ioctl(r2,1074310803,&kvm_vapic_addr_1);
        return 0;
}

Consider adding a bound-check in if-condition as following code :

if (start < slots->used_slots && gfn >= memslots[start].base_gfn &&
            gfn < memslots[start].base_gfn + memslots[start].npages) {
                atomic_set(&slots->lru_slot, start);
                return &memslots[start];
}

-- 
Hao Sun