Skip to contentSkip to conversation

x86/ucode/amd: Fix more potential buffer overruns with microcode parsing

PATCH In review Andrew Cooper 2 participants 3 messages Mar 28, 2020 – Mar 30, 2020

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 x86/ucode/amd: Fix more potential buffer overruns with microcode parsing

cpu_request_microcode() doesn't know the buffer is at least 4 bytes long before inspecting UCODE_MAGIC.

install_equiv_cpu_table() doesn't know the boundary of the buffer it is interpreting as an equivalency table. This case was clearly observed at one point in the past, given the subsequent overrun detection, but without comprehending that the damage was already done.

Make the logic consistent with container_fast_forward() and pass size_left in to install_equiv_cpu_table().

---
CC: Jan Beulich <[email protected]>
CC: Wei Liu <[email protected]>
CC: Roger Pau Monné <[email protected]>

---
 xen/arch/x86/cpu/microcode/amd.c | 27 +++++++++++++--------------
 1 file changed, 13 insertions(+), 14 deletions(-)
xen/arch/x86/cpu/microcode/amd.c +13 -14
index 6bf3a054d3..796745e928 100644
@@ -303,11 +303,20 @@ static int get_ucode_from_buffer_amd(
 static int install_equiv_cpu_table(
     struct microcode_amd *mc_amd,
     const void *data,
+    size_t size_left,
     size_t *offset)
 {
-    const struct mpbhdr *mpbuf = data + *offset + 4;
+    const struct mpbhdr *mpbuf;
     const struct equiv_cpu_entry *eq;
 
+    if ( size_left < (sizeof(*mpbuf) + 4) ||
+         (mpbuf = data + *offset + 4,
+          size_left - sizeof(*mpbuf) - 4 < mpbuf->len) )
+    {
+        printk(XENLOG_WARNING "microcode: No space for equivalent cpu table\n");
+        return -EINVAL;
+    }
+
     *offset += mpbuf->len + CONT_HDR_SIZE;	/* add header length */
 
     if ( mpbuf->type != UCODE_EQUIV_CPU_TABLE_TYPE )
@@ -417,7 +426,8 @@ static struct microcode_patch *cpu_request_microcode(const void *buf,
 
     current_cpu_id = cpuid_eax(0x00000001);
 
-    if ( *(const uint32_t *)buf != UCODE_MAGIC )
+    if ( bufsize < 4 ||
+         *(const uint32_t *)buf != UCODE_MAGIC )
     {
         printk(KERN_ERR "microcode: Wrong microcode patch file magic\n");
         error = -EINVAL;
@@ -447,24 +457,13 @@ static struct microcode_patch *cpu_request_microcode(const void *buf,
      */
     while ( offset < bufsize )
     {
-        error = install_equiv_cpu_table(mc_amd, buf, &offset);
+        error = install_equiv_cpu_table(mc_amd, buf, bufsize - offset, &offset);
         if ( error )
         {
             printk(KERN_ERR "microcode: installing equivalent cpu table failed\n");
             break;
         }
 
-        /*
-         * Could happen as we advance 'offset' early
-         * in install_equiv_cpu_table
-         */
-        if ( offset > bufsize )
-        {
-            printk(KERN_ERR "microcode: Microcode buffer overrun\n");
-            error = -EINVAL;
-            break;
-        }
-
         if ( find_equiv_cpu_id(mc_amd->equiv_cpu_table, current_cpu_id,
                                &equiv_cpu_id) )
             break;
Signature · 1 line
2.11.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 6 lines · Andrew Cooper
xen/arch/x86/cpu/microcode/amd.c +0 -0
@@ -303,11 +303,20 @@ static int get_ucode_from_buffer_amd(
 static int install_equiv_cpu_table(
     struct microcode_amd *mc_amd,
     const void *data,

xen/arch/x86/cpu/microcode/amd.c static int get_ucode_from_buffer_amd( · line 314

Quoted 10 lines
xen/arch/x86/cpu/microcode/amd.c +5 -1
+    size_t size_left,
     size_t *offset)
 {
-    const struct mpbhdr *mpbuf = data + *offset + 4;
+    const struct mpbhdr *mpbuf;
     const struct equiv_cpu_entry *eq;
 
+    if ( size_left < (sizeof(*mpbuf) + 4) ||
+         (mpbuf = data + *offset + 4,
+          size_left - sizeof(*mpbuf) - 4 < mpbuf->len) )

This proliferation of literal 4 made me look into where this 4 is coming from and why it's here. Afaict it's covering for UCODE_MAGIC. There's no good sizeof() that could be used instead, so how about getting rid of this addend altogether by setting offset to sizeof(uint32_t) near

    if ( *(const uint32_t *)buf != UCODE_MAGIC )

in cpu_request_microcode() and adding sizeof(header[0]) into *offset near

        if ( header[0] == UCODE_MAGIC &&
             header[1] == UCODE_EQUIV_CPU_TABLE_TYPE )

in container_fast_forward()? (The code following the big, 7- bullet-point comment in cpu_request_microcode() may also need adjustment, but I wonder if we couldn't better get rid of it altogether.)

However, since the change as is looks correct to me and doesn't make the situation much worse, I'm not objecting to you also putting it in as is, in which case

Jan

#2 lore raw
Quoted 14 lines · Jan Beulich
Quoted 13 lines · Andrew Cooper
xen/arch/x86/cpu/microcode/amd.c +2 -1
@@ -303,11 +303,20 @@ static int get_ucode_from_buffer_amd(
 static int install_equiv_cpu_table(
     struct microcode_amd *mc_amd,
     const void *data,
+    size_t size_left,
     size_t *offset)
 {
-    const struct mpbhdr *mpbuf = data + *offset + 4;
+    const struct mpbhdr *mpbuf;
     const struct equiv_cpu_entry *eq;
 
Quoted 6 lines

xen/arch/x86/cpu/microcode/amd.c static int get_ucode_from_buffer_amd( · line 314

Quoted 3 lines
xen/arch/x86/cpu/microcode/amd.c +3 -0
+    if ( size_left < (sizeof(*mpbuf) + 4) ||
+         (mpbuf = data + *offset + 4,
+          size_left - sizeof(*mpbuf) - 4 < mpbuf->len) )

This proliferation of literal 4 made me look into where this 4 is coming from and why it's here. Afaict it's covering for UCODE_MAGIC.

Correct

Quoted 6 lines

There's no good sizeof() that could be used instead, so how about getting rid of this addend altogether by setting offset to sizeof(uint32_t) near

    if ( *(const uint32_t *)buf != UCODE_MAGIC )
Quoted 7 lines

in cpu_request_microcode() and adding sizeof(header[0]) into *offset near

        if ( header[0] == UCODE_MAGIC &&
             header[1] == UCODE_EQUIV_CPU_TABLE_TYPE )

in container_fast_forward()?

I am in the process of rewriting all of this.  I'm not entirely convinced of the correctness of cpu_request_microcode() in the first place, but I also can't find a concrete problem with it.

Quoted 4 lines

(The code following the big, 7-
bullet-point comment in cpu_request_microcode() may also need adjustment, but I wonder if we couldn't better get rid of it altogether.)

I'm unconvinced by what is described there. It honestly seems easier to scan to the end of all containers.

Quoted 5 lines

However, since the change as is looks correct to me and doesn't make the situation much worse, I'm not objecting to you also putting it in as is, in which case

Thanks.  I'm tempted to keep it as-is because I think it is rather more amenable to backport.

However, given the rate of finding issues, I'm going to hold off on committing this until I've completed the rewrite, and am fairly sure there are no further issues lurking.

~Andrew