On Tue, 2019-07-30 at 11:51 +0200, Pavel Machek wrote: > Hi! > > Code does strcmps, but does not actually check count. So AFAICT > writing "host-I-don't-want-I-need-peripheral" into the file will > succeed and turn it into host mode. I suspect the intent is to work without a trailing "\n", but you're right that this parsing is rather lax. > Also data beyond count in buf are going to be accessed. All attribute writes should be going through kernfs_fop_write(), which writes a null terminator at buf[count] before calling the write/store function. (I don't know where that'ss documented though.) So I think this is safe. Ben. > static ssize_t role_store(struct device *dev, struct device_attribute *attr, > const char *buf, size_t count) > { > struct rcar_gen3_chan *ch = dev_get_drvdata(dev); > bool is_b_device; > enum phy_mode cur_mode, new_mode; > > if (!ch->is_otg_channel || !rcar_gen3_is_any_rphy_initialized(ch)) > return -EIO; > > /* FIXME, this is wrong */ > if (!strncmp(buf, "host", strlen("host"))) > new_mode = PHY_MODE_USB_HOST; > else if (!strncmp(buf, "peripheral", strlen("peripheral"))) > new_mode = PHY_MODE_USB_DEVICE; > else > return -EINVAL; > > > _______________________________________________ > cip-dev mailing list > cip-dev@xxxxxxxxxxxxxxxxxxxxx > https://lists.cip-project.org/mailman/listinfo/cip-dev -- Ben Hutchings, Software Developer Codethink Ltd https://www.codethink.co.uk/ Dale House, 35 Dale Street Manchester, M1 2HF, United Kingdom