Fix interval_tree_iter_first() to check root node value

Posted by Helge Deller 1 year, 3 months ago

Fix a crash in qemu-user when running

    cat /proc/self/maps

in a chroot, where /proc isn't mounted.

The problem was introduced by commit 3ce3dd8ca965 ("util/selfmap:
Rewrite using qemu/interval-tree.h") where in open_self_maps_1() the
function read_self_maps() is called and which returns NULL if it can't
read the hosts /proc/self/maps file. Afterwards that NULL is fed into
interval_tree_iter_first() which doesn't check if the root node is NULL.

Fix it by adding a check if root is NULL and return NULL in that case.

Signed-off-by: Helge Deller <deller@gmx.de>
Fixes: 3ce3dd8ca965 ("util/selfmap: Rewrite using qemu/interval-tree.h")

diff --git a/util/interval-tree.c b/util/interval-tree.c
index f2866aa7d3..53465182e6 100644
--- a/util/interval-tree.c
+++ b/util/interval-tree.c
@@ -797,7 +797,7 @@ IntervalTreeNode *interval_tree_iter_first(IntervalTreeRoot *root,
 {
     IntervalTreeNode *node, *leftmost;

-    if (!root->rb_root.rb_node) {
+    if (!root || !root->rb_root.rb_node) {
         return NULL;
     }

Re: Fix interval_tree_iter_first() to check root node value

Posted by Richard Henderson 1 year, 3 months ago

On 8/9/23 08:11, Helge Deller wrote:
> Fix a crash in qemu-user when running
> 
>      cat /proc/self/maps
> 
> in a chroot, where /proc isn't mounted.
> 
> The problem was introduced by commit 3ce3dd8ca965 ("util/selfmap:
> Rewrite using qemu/interval-tree.h") where in open_self_maps_1() the
> function read_self_maps() is called and which returns NULL if it can't
> read the hosts /proc/self/maps file. Afterwards that NULL is fed into
> interval_tree_iter_first() which doesn't check if the root node is NULL.
> 
> Fix it by adding a check if root is NULL and return NULL in that case.
> 
> Signed-off-by: Helge Deller <deller@gmx.de>
> Fixes: 3ce3dd8ca965 ("util/selfmap: Rewrite using qemu/interval-tree.h")
> 
> diff --git a/util/interval-tree.c b/util/interval-tree.c
> index f2866aa7d3..53465182e6 100644
> --- a/util/interval-tree.c
> +++ b/util/interval-tree.c
> @@ -797,7 +797,7 @@ IntervalTreeNode *interval_tree_iter_first(IntervalTreeRoot *root,
>   {
>       IntervalTreeNode *node, *leftmost;
> 
> -    if (!root->rb_root.rb_node) {
> +    if (!root || !root->rb_root.rb_node) {


I guess this is good enough for 8.1.  Before the conversion to interval-tree we would also 
emit nothing.

I've already done a rewrite for 8.2, and I noticed this problem.  There I emit what 
mapping information that I have, which is everything except for the device+path data.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~

Re: Fix interval_tree_iter_first() to check root node value

Posted by Helge Deller 1 year, 3 months ago

On 8/9/23 17:23, Richard Henderson wrote:
> On 8/9/23 08:11, Helge Deller wrote:
>> Fix a crash in qemu-user when running
>>
>>      cat /proc/self/maps
>>
>> in a chroot, where /proc isn't mounted.
>>
>> The problem was introduced by commit 3ce3dd8ca965 ("util/selfmap:
>> Rewrite using qemu/interval-tree.h") where in open_self_maps_1() the
>> function read_self_maps() is called and which returns NULL if it can't
>> read the hosts /proc/self/maps file. Afterwards that NULL is fed into
>> interval_tree_iter_first() which doesn't check if the root node is NULL.
>>
>> Fix it by adding a check if root is NULL and return NULL in that case.
>>
>> Signed-off-by: Helge Deller <deller@gmx.de>
>> Fixes: 3ce3dd8ca965 ("util/selfmap: Rewrite using qemu/interval-tree.h")
>>
>> diff --git a/util/interval-tree.c b/util/interval-tree.c
>> index f2866aa7d3..53465182e6 100644
>> --- a/util/interval-tree.c
>> +++ b/util/interval-tree.c
>> @@ -797,7 +797,7 @@ IntervalTreeNode *interval_tree_iter_first(IntervalTreeRoot *root,
>>   {
>>       IntervalTreeNode *node, *leftmost;
>>
>> -    if (!root->rb_root.rb_node) {
>> +    if (!root || !root->rb_root.rb_node) {
>
>
> I guess this is good enough for 8.1.  Before the conversion to interval-tree we would also emit nothing.

Yes and yes.

> I've already done a rewrite for 8.2, and I noticed this problem.
> There I emit what mapping information that I have, which is
> everything except for the device+path data.

nice.

> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

Shall I send a pull request?
If so, is it OK that I include this patch in the pull-request as well?
   linux-user: Fix openat() emulation to correctly detect accesses to /proc
   https://lists.nongnu.org/archive/html/qemu-devel/2023-08/msg00165.html
which already has been R-b: Daniel P. Berrangé

Helge

Re: Fix interval_tree_iter_first() to check root node value

Posted by Richard Henderson 1 year, 3 months ago

On 8/9/23 08:53, Helge Deller wrote:
> On 8/9/23 17:23, Richard Henderson wrote:
>> On 8/9/23 08:11, Helge Deller wrote:
>>> Fix a crash in qemu-user when running
>>>
>>>      cat /proc/self/maps
>>>
>>> in a chroot, where /proc isn't mounted.
>>>
>>> The problem was introduced by commit 3ce3dd8ca965 ("util/selfmap:
>>> Rewrite using qemu/interval-tree.h") where in open_self_maps_1() the
>>> function read_self_maps() is called and which returns NULL if it can't
>>> read the hosts /proc/self/maps file. Afterwards that NULL is fed into
>>> interval_tree_iter_first() which doesn't check if the root node is NULL.
>>>
>>> Fix it by adding a check if root is NULL and return NULL in that case.
>>>
>>> Signed-off-by: Helge Deller <deller@gmx.de>
>>> Fixes: 3ce3dd8ca965 ("util/selfmap: Rewrite using qemu/interval-tree.h")
>>>
>>> diff --git a/util/interval-tree.c b/util/interval-tree.c
>>> index f2866aa7d3..53465182e6 100644
>>> --- a/util/interval-tree.c
>>> +++ b/util/interval-tree.c
>>> @@ -797,7 +797,7 @@ IntervalTreeNode *interval_tree_iter_first(IntervalTreeRoot *root,
>>>   {
>>>       IntervalTreeNode *node, *leftmost;
>>>
>>> -    if (!root->rb_root.rb_node) {
>>> +    if (!root || !root->rb_root.rb_node) {
>>
>>
>> I guess this is good enough for 8.1.  Before the conversion to interval-tree we would 
>> also emit nothing.
> 
> Yes and yes.
> 
>> I've already done a rewrite for 8.2, and I noticed this problem.
>> There I emit what mapping information that I have, which is
>> everything except for the device+path data.
> 
> nice.
> 
>> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> 
> Shall I send a pull request?
> If so, is it OK that I include this patch in the pull-request as well?
>    linux-user: Fix openat() emulation to correctly detect accesses to /proc
>    https://lists.nongnu.org/archive/html/qemu-devel/2023-08/msg00165.html
> which already has been R-b: Daniel P. Berrangé

I can pick them both up -- I have other linux-user patches to send.


r~