Hi, Philippe
Thanks for your quick review:)
On 05/18/2017 01:22 AM, Philippe Mathieu-Daudé wrote:
Hi Mao,
On 05/17/2017 08:12 AM, Mao Zhongyi wrote:
pci_rocker_init() leaks a World when the name more than 9 chars,
then return a negative value directly, doesn't make a correct
cleanup. So add a new goto label to fix it.
Cc: jasow...@redhat.com
Cc: j...@resnulli.us
Cc: f4...@amsat.org
Cc: arm...@redhat.com
Signed-off-by: Mao Zhongyi <maozy.f...@cn.fujitsu.com>
---
hw/net/rocker/rocker.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/hw/net/rocker/rocker.c b/hw/net/rocker/rocker.c
index b2b6dc7..a382a6f 100644
--- a/hw/net/rocker/rocker.c
+++ b/hw/net/rocker/rocker.c
@@ -1371,7 +1371,8 @@ static int pci_rocker_init(PCIDevice *dev)
fprintf(stderr,
While here can you update fprintf -> qemu_log_mask?
In the patch 3, fprintf() has been updated to err_setg() which output error
message
for more user-friendly. I'm not very clear why convert it to qemu_log_mask().
Could
you give me more hints about this?
Meanwhile, updating to qemu_log_mask() results in a segmentation fault. Because
the
World memory was freed for 2 times. The first time is when the name more than 9
chars,
It was freed by the goto label "err_name_too_long" in the pci_rocker_realize(),
the
second is freed by qemu_system_reset() which called in the main().
So I think convert it to Error maybe a little better. What do you think?
Thanks
Mao
"rocker: name too long; please shorten to at most %d chars\n",
MAX_ROCKER_NAME_LEN);
- return -EINVAL;
+ err = -EINVAL;
+ goto err_name_too_long;
}
if (memcmp(&r->fp_start_macaddr, &zero, sizeof(zero)) == 0) {
@@ -1430,6 +1431,7 @@ static int pci_rocker_init(PCIDevice *dev)
return 0;
+err_name_too_long:
err_duplicate:
rocker_msix_uninit(r);
err_msix_init:
