Dan Carpenter
2021-Apr-21 13:24 UTC
[Nouveau] [bug report] drm/nouveau/hwmon: Remove old code, add .write/.read operations
Hello Oscar Salvador,
The patch bfb96e4c344e: "drm/nouveau/hwmon: Remove old code, add
.write/.read operations" from May 18, 2017, leads to the following
static checker warning:
drivers/gpu/drm/nouveau/nouveau_hwmon.c:507 nouveau_in_read()
warn: check sign expansion for '-19'
drivers/gpu/drm/nouveau/nouveau_hwmon.c:510 nouveau_in_read()
warn: check sign expansion for '-19'
drivers/gpu/drm/nouveau/nouveau_hwmon.c
488 static int
489 nouveau_in_read(struct device *dev, u32 attr, int channel, long *val)
490 {
491 struct drm_device *drm_dev = dev_get_drvdata(dev);
492 struct nouveau_drm *drm = nouveau_drm(drm_dev);
493 struct nvkm_volt *volt = nvxx_volt(&drm->client.device);
494 int ret;
495
496 if (!volt)
497 return -EOPNOTSUPP;
498
499 switch (attr) {
500 case hwmon_in_input:
501 if (drm_dev->switch_power_state !=
DRM_SWITCH_POWER_ON)
502 return -EINVAL;
503 ret = nvkm_volt_get(volt);
504 *val = ret < 0 ? ret : (ret / 1000);
505 break;
506 case hwmon_in_min:
507 *val = volt->min_uv > 0 ? (volt->min_uv / 1000)
: -ENODEV;
This is trying to set "*val = -ENODEV" but because
"volt->min_uv" is
unsigned int it actually sets it to "*val = (unsigned int)-ENODEV".
It's weird to me that this code doesn't return -ENODEV instead of
setting *val to it.
508 break;
509 case hwmon_in_max:
510 *val = volt->max_uv > 0 ? (volt->max_uv / 1000)
: -ENODEV;
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
511 break;
512 default:
513 return -EOPNOTSUPP;
514 }
515
516 return 0;
517 }
regards,
dan carpenter
Karol Herbst
2021-Apr-21 16:04 UTC
[Nouveau] [bug report] drm/nouveau/hwmon: Remove old code, add .write/.read operations
On Wed, Apr 21, 2021 at 3:57 PM Dan Carpenter <dan.carpenter at oracle.com> wrote:> > Hello Oscar Salvador, > > The patch bfb96e4c344e: "drm/nouveau/hwmon: Remove old code, add > .write/.read operations" from May 18, 2017, leads to the following > static checker warning: > > drivers/gpu/drm/nouveau/nouveau_hwmon.c:507 nouveau_in_read() > warn: check sign expansion for '-19' > drivers/gpu/drm/nouveau/nouveau_hwmon.c:510 nouveau_in_read() > warn: check sign expansion for '-19' > > drivers/gpu/drm/nouveau/nouveau_hwmon.c > 488 static int > 489 nouveau_in_read(struct device *dev, u32 attr, int channel, long *val) > 490 { > 491 struct drm_device *drm_dev = dev_get_drvdata(dev); > 492 struct nouveau_drm *drm = nouveau_drm(drm_dev); > 493 struct nvkm_volt *volt = nvxx_volt(&drm->client.device); > 494 int ret; > 495 > 496 if (!volt) > 497 return -EOPNOTSUPP; > 498 > 499 switch (attr) { > 500 case hwmon_in_input: > 501 if (drm_dev->switch_power_state != DRM_SWITCH_POWER_ON) > 502 return -EINVAL; > 503 ret = nvkm_volt_get(volt); > 504 *val = ret < 0 ? ret : (ret / 1000); > 505 break; > 506 case hwmon_in_min: > 507 *val = volt->min_uv > 0 ? (volt->min_uv / 1000) : -ENODEV; > > This is trying to set "*val = -ENODEV" but because "volt->min_uv" is > unsigned int it actually sets it to "*val = (unsigned int)-ENODEV". > > It's weird to me that this code doesn't return -ENODEV instead of > setting *val to it. >ohh.. that might actually be a left over from the conversion we've done in the past. Thanks for pointing it out. Do you want to write the patch as well?> 508 break; > 509 case hwmon_in_max: > 510 *val = volt->max_uv > 0 ? (volt->max_uv / 1000) : -ENODEV; > ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ > > 511 break; > 512 default: > 513 return -EOPNOTSUPP; > 514 } > 515 > 516 return 0; > 517 } > > regards, > dan carpenter > _______________________________________________ > Nouveau mailing list > Nouveau at lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/nouveau >