[PATCH v2 4/4] remoteproc: debugfs: Add ability to boot remote processor using debugfs
flat view
From: Nathan Lynch <hidden>
Date: 2015-08-28 15:24:43
Also in:
linux-devicetree, lkml
On 08/28/2015 05:31 AM, Lee Jones wrote:
quoted hunk ↗ jump to hunk
+static ssize_t rproc_state_write(struct file *filp, const char __user *userbuf, + size_t count, loff_t *ppos) +{ + struct rproc *rproc = filp->private_data; + char buf[2]; + int ret; + + ret = copy_from_user(buf, userbuf, 1); + if (ret) + return -EFAULT; + + switch (buf[0]) { + case '0': + rproc_shutdown(rproc); + break; + case '1': + ret = rproc_boot(rproc); + if (ret) + dev_warn(&rproc->dev, "Boot failed: %d\n", ret); + break; + default: + dev_err(&rproc->dev, "Unrecognised option: %x\n", buf[1]); + return -EINVAL;
This prints uninitialized kernel stack contents instead of what was copied from user space. Is the dev_err statement really necessary anyway?
quoted hunk ↗ jump to hunk
+ } + + return count; +}
If rproc_boot fails, that should be reflected in the syscall result. This interface is essentially exposing the remoteproc->power refcount to user space; is that okay? Seems like it makes it easy to underflow remoteproc->power through successive shutdown calls. The other debugfs interface in remoteproc that has a write method (recovery) accepts more expressive string commands as opposed to 0/1. It would be more consistent for this interface to take commands such as "boot" and "shutdown" IMO.