Skip to contentSkip to conversation

asix: fix uninit-value in asix_mdio_read()

bug report Dan Carpenter 2 participants 2 messages Jan 6, 2022 Kernel janitors

lore
Share & tools
lore
#0 lore raw

Hello Pavel Skripkin,

The patch 8035b1a2a37a: "asix: fix uninit-value in asix_mdio_read()" from Dec 21, 2021, leads to the following Smatch static checker warning:

drivers/net/usb/asix_common.c:82 asix_check_host_enable()
warn: 'ret' possible negative type promoted to high

drivers/net/usb/asix_common.c

    68 static int asix_check_host_enable(struct usbnet *dev, int in_pm)
    69 {
    70         int i, ret;
    71         u8 smsr;
    72 
    73         for (i = 0; i < AX_HOST_EN_RETRIES; ++i) {
    74                 ret = asix_set_sw_mii(dev, in_pm);
    75                 if (ret == -ENODEV || ret == -ETIMEDOUT)
    76                         break;
    77                 usleep_range(1000, 1100);
    78                 ret = asix_read_cmd(dev, AX_CMD_STATMNGSTS_REG,
    79                                     0, 0, 1, &smsr, in_pm);
    80                 if (ret == -ENODEV)
    81                         break;

--> 82 else if (ret < sizeof(smsr))

This has to be: if (ret < 0 || ret < sizeof(smsr)) { but even better would be to fix asix_read_cmd() to not allow partial reads. It should print the netdev_warn() for partial reads but right now it doesn't.

    83                         continue;
    84                 else if (smsr & AX_HOST_EN)
    85                         break;
    86         }
    87 
    88         return i >= AX_HOST_EN_RETRIES ? -ETIMEDOUT : ret;
    89 }

regards,
dan carpenter

#1 lore raw

Hi Dan,

Quoted 24 lines · Dan Carpenter

Hello Pavel Skripkin,

The patch 8035b1a2a37a: "asix: fix uninit-value in asix_mdio_read()" from Dec 21, 2021, leads to the following Smatch static checker warning:

drivers/net/usb/asix_common.c:82 asix_check_host_enable()
warn: 'ret' possible negative type promoted to high

drivers/net/usb/asix_common.c

     68 static int asix_check_host_enable(struct usbnet *dev, int in_pm)
     69 {
     70         int i, ret;
     71         u8 smsr;
     72
     73         for (i = 0; i < AX_HOST_EN_RETRIES; ++i) {
     74                 ret = asix_set_sw_mii(dev, in_pm);
     75                 if (ret == -ENODEV || ret == -ETIMEDOUT)
     76                         break;
     77                 usleep_range(1000, 1100);
     78                 ret = asix_read_cmd(dev, AX_CMD_STATMNGSTS_REG,
     79                                     0, 0, 1, &smsr, in_pm);
     80                 if (ret == -ENODEV)
     81                         break;
Quoted 4 lines

--> 82 else if (ret < sizeof(smsr))

This has to be: if (ret < 0 || ret < sizeof(smsr)) { but even better would be to fix asix_read_cmd() to not allow partial reads. It should

Thanks for reporting this. It's indeed a bug.

I sent the fix yesterday, that disallows partial reads inside asix_read_cmd()

Please see

https://lore.kernel.org/all/[email protected]/

With regards,
Pavel Skripkin