util/qemu-openpty.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-)
As we can see from the following function call stack, the amaster and the aslave
cannot be NULL: char_pty_open() -> qemu_openpty_raw() -> openpty().
In addition, the amaster and the aslave has been dereferenced at the beginning
of the openpty(). So the checks on amaster and aslave in the openpty() are redundant.
Reported-by: Euler Robot <euler.robot@huawei.com>
Signed-off-by: Alex Chen <alex.chen@huawei.com>
---
util/qemu-openpty.c | 7 +++----
1 file changed, 3 insertions(+), 4 deletions(-)
diff --git a/util/qemu-openpty.c b/util/qemu-openpty.c
index eb17f5b0bc..427f43a769 100644
--- a/util/qemu-openpty.c
+++ b/util/qemu-openpty.c
@@ -80,10 +80,9 @@ static int openpty(int *amaster, int *aslave, char *name,
(termp != NULL && tcgetattr(sfd, termp) < 0))
goto err;
- if (amaster)
- *amaster = mfd;
- if (aslave)
- *aslave = sfd;
+ *amaster = mfd;
+ *aslave = sfd;
+
if (winp)
ioctl(sfd, TIOCSWINSZ, winp);
--
2.19.1
On Sat, 31 Oct 2020 at 11:04, AlexChen <alex.chen@huawei.com> wrote: > > As we can see from the following function call stack, the amaster and the aslave > cannot be NULL: char_pty_open() -> qemu_openpty_raw() -> openpty(). > In addition, the amaster and the aslave has been dereferenced at the beginning > of the openpty(). So the checks on amaster and aslave in the openpty() are redundant. > > Reported-by: Euler Robot <euler.robot@huawei.com> > Signed-off-by: Alex Chen <alex.chen@huawei.com> This function is trying to match the BSD/glibc openpty() function, so the thing to check here is not QEMU's specific current usage but the API specification for openpty(): https://www.gnu.org/software/libc/manual/html_node/Pseudo_002dTerminal-Pairs.html https://www.freebsd.org/cgi/man.cgi?query=openpty The spec says that name, termp and winp can all be NULL, but it doesn't say this for amaster and aslave, so indeed the change in this patch is the correct one. > --- > util/qemu-openpty.c | 7 +++---- > 1 file changed, 3 insertions(+), 4 deletions(-) > > diff --git a/util/qemu-openpty.c b/util/qemu-openpty.c > index eb17f5b0bc..427f43a769 100644 > --- a/util/qemu-openpty.c > +++ b/util/qemu-openpty.c > @@ -80,10 +80,9 @@ static int openpty(int *amaster, int *aslave, char *name, > (termp != NULL && tcgetattr(sfd, termp) < 0)) > goto err; > > - if (amaster) > - *amaster = mfd; > - if (aslave) > - *aslave = sfd; > + *amaster = mfd; > + *aslave = sfd; > + > if (winp) > ioctl(sfd, TIOCSWINSZ, winp); Reviewed-by: Peter Maydell <peter.maydell@linaro.org> though you might like to mention in the commit message that the openpty() API doesn't allow NULL amaster or aslave arguments. thanks -- PMM
On 2020/10/31 23:21, Peter Maydell wrote:
> On Sat, 31 Oct 2020 at 11:04, AlexChen <alex.chen@huawei.com> wrote:
>>
>> As we can see from the following function call stack, the amaster and the aslave
>> cannot be NULL: char_pty_open() -> qemu_openpty_raw() -> openpty().
>> In addition, the amaster and the aslave has been dereferenced at the beginning
>> of the openpty(). So the checks on amaster and aslave in the openpty() are redundant.
>>
>> Reported-by: Euler Robot <euler.robot@huawei.com>
>> Signed-off-by: Alex Chen <alex.chen@huawei.com>
>
> This function is trying to match the BSD/glibc openpty()
> function, so the thing to check here is not QEMU's specific
> current usage but the API specification for openpty():
> https://www.gnu.org/software/libc/manual/html_node/Pseudo_002dTerminal-Pairs.html
> https://www.freebsd.org/cgi/man.cgi?query=openpty
>
> The spec says that name, termp and winp can all be
> NULL, but it doesn't say this for amaster and aslave,
> so indeed the change in this patch is the correct one.
>
>> ---
>> util/qemu-openpty.c | 7 +++----
>> 1 file changed, 3 insertions(+), 4 deletions(-)
>>
>> diff --git a/util/qemu-openpty.c b/util/qemu-openpty.c
>> index eb17f5b0bc..427f43a769 100644
>> --- a/util/qemu-openpty.c
>> +++ b/util/qemu-openpty.c
>> @@ -80,10 +80,9 @@ static int openpty(int *amaster, int *aslave, char *name,
>> (termp != NULL && tcgetattr(sfd, termp) < 0))
>> goto err;
>>
>> - if (amaster)
>> - *amaster = mfd;
>> - if (aslave)
>> - *aslave = sfd;
>> + *amaster = mfd;
>> + *aslave = sfd;
>> +
>> if (winp)
>> ioctl(sfd, TIOCSWINSZ, winp);
>
> Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
>
> though you might like to mention in the commit message that
> the openpty() API doesn't allow NULL amaster or aslave
> arguments.
>
Thanks for your review, I will add this description to my commit message in my patch V2.
In addition, since the amaster and the aslave are not allow to be NULL,
do we need to check that the amaster and the aslave are NULL in the beginning of the openpty()?
such as this modification:
diff --git a/util/qemu-openpty.c b/util/qemu-openpty.c
index eb17f5b0bc..1aadd39395 100644
--- a/util/qemu-openpty.c
+++ b/util/qemu-openpty.c
@@ -61,6 +61,9 @@ static int openpty(int *amaster, int *aslave, char *name,
const char *slave;
int mfd = -1, sfd = -1;
+ if (!amaster || !aslave)
+ goto err;
+
*amaster = *aslave = -1;
mfd = open("/dev/ptmx", O_RDWR | O_NOCTTY);
@@ -80,10 +83,9 @@ static int openpty(int *amaster, int *aslave, char *name,
(termp != NULL && tcgetattr(sfd, termp) < 0))
goto err;
- if (amaster)
- *amaster = mfd;
- if (aslave)
- *aslave = sfd;
+ *amaster = mfd;
+ *aslave = sfd;
+
if (winp)
ioctl(sfd, TIOCSWINSZ, winp);
@@ -92,7 +94,8 @@ static int openpty(int *amaster, int *aslave, char *name,
err:
if (sfd != -1)
close(sfd);
- close(mfd);
+ if (mfd != -1)
+ close(mfd);
return -1;
}
#endif
--
2.19.1
Thanks,
Alex
On Mon, 2 Nov 2020 at 05:12, AlexChen <alex.chen@huawei.com> wrote: > Thanks for your review, I will add this description to my commit message in my patch V2. > In addition, since the amaster and the aslave are not allow to be NULL, > do we need to check that the amaster and the aslave are NULL in the beginning of the openpty()? No, we can just assume it. Our coding style doesn't mandate that level of defensive-coding against caller errors. thanks -- PMM
© 2016 - 2024 Red Hat, Inc.