On 03/06/14 17:37, Mathieu Poirier wrote: >>> +static ssize_t debugfs_status_read(struct file *file, char __user >>> *user_buf, >>> + size_t count, loff_t *ppos) >>> +{ >>> + ssize_t ret; >>> + uint32_t val; >>> + unsigned long flags; >>> + char *buf = kmalloc(PAGE_SIZE, GFP_KERNEL); >>> + struct etm_drvdata *drvdata = file->private_data; >>> + >>> + if (!buf) >>> + return -ENOMEM; >>> + >>> + ret = clk_prepare_enable(drvdata->clk); >>> + if (ret) >>> + goto out; >>> + >>> + spin_lock_irqsave(&drvdata->spinlock, flags); >>> + >>> + ETM_UNLOCK(drvdata); >>> + val = etm_readl(drvdata, ETMCCR); >>> + ret += sprintf(buf, "ETMCCR: 0x%08x\n", val); >>> + val = etm_readl(drvdata, ETMCCER); >>> + ret += sprintf(buf + ret, "ETMCCER: 0x%08x\n", val); >>> + val = etm_readl(drvdata, ETMSCR); >>> + ret += sprintf(buf + ret, "ETMSCR: 0x%08x\n", val); >>> + val = etm_readl(drvdata, ETMIDR); >>> + ret += sprintf(buf + ret, "ETMIDR: 0x%08x\n", val); >>> + val = etm_readl(drvdata, ETMCR); >>> + ret += sprintf(buf + ret, "ETMCR: 0x%08x\n", val); >>> + val = etm_readl(drvdata, ETMTEEVR); >>> + ret += sprintf(buf + ret, "Enable event: 0x%08x\n", val); >>> + val = etm_readl(drvdata, ETMTSSCR); >>> + ret += sprintf(buf + ret, "Enable start/stop: 0x%08x\n", val); >>> + ret += sprintf(buf + ret, >>> + "Enable control: CR1 0x%08x CR2 0x%08x\n", >>> + etm_readl(drvdata, ETMTECR1), >>> + etm_readl(drvdata, ETMTECR2)); >>> + >>> + ETM_LOCK(drvdata); >>> + >>> + spin_unlock_irqrestore(&drvdata->spinlock, flags); >>> + clk_disable_unprepare(drvdata->clk); >>> + >>> + ret = simple_read_from_buffer(user_buf, count, ppos, buf, ret); >>> +out: >>> + kfree(buf); >>> + return ret; >>> +} >> >> Really not sure whether this should be in the read method. If we don't >> read the file in one go the spin_lock() we'll not get a cohesive set of >> registers. > > I get your point but since there is a possibility (even very remove) > that any of these registers can be changed between the two read > operations, the only reasonable solution I see is to return an error > if (ret > size). What your opinion on that?
I'd prefer that we simply copy the approach used by simple_attr_read(). Daniel. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/