/* riowd.c - driver for hw watchdog inside Super I/O of RIO * * Copyright (C) 2001, 2008 David S. Miller (davem@davemloft.net) */ #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt #include #include #include #include #include #include #include #include #include #include #include #include /* RIO uses the NatSemi Super I/O power management logical device * as its' watchdog. * * When the watchdog triggers, it asserts a line to the BBC (Boot Bus * Controller) of the machine. The BBC can only be configured to * trigger a power-on reset when the signal is asserted. The BBC * can be configured to ignore the signal entirely as well. * * The only Super I/O device register we care about is at index * 0x05 (WDTO_INDEX) which is the watchdog time-out in minutes (1-255). * If set to zero, this disables the watchdog. When set, the system * must periodically (before watchdog expires) clear (set to zero) and * re-set the watchdog else it will trigger. * * There are two other indexed watchdog registers inside this Super I/O * logical device, but they are unused. The first, at index 0x06 is * the watchdog control and can be used to make the watchdog timer re-set * when the PS/2 mouse or serial lines show activity. The second, at * index 0x07 is merely a sampling of the line from the watchdog to the * BBC. * * The watchdog device generates no interrupts. */ MODULE_AUTHOR("David S. Miller "); MODULE_DESCRIPTION("Hardware watchdog driver for Sun RIO"); MODULE_SUPPORTED_DEVICE("watchdog"); MODULE_LICENSE("GPL"); #define DRIVER_NAME "riowd" #define PFX DRIVER_NAME ": " struct riowd { void __iomem *regs; spinlock_t lock; }; static struct riowd *riowd_device; #define WDTO_INDEX 0x05 static int riowd_timeout = 1; /* in minutes */ module_param(riowd_timeout, int, 0); MODULE_PARM_DESC(riowd_timeout, "Watchdog timeout in minutes"); static void riowd_writereg(struct riowd *p, u8 val, int index) { unsigned long flags; spin_lock_irqsave(&p->lock, flags); writeb(index, p->regs + 0); writeb(val, p->regs + 1); spin_unlock_irqrestore(&p->lock, flags); } static int riowd_open(struct inode *inode, struct file *filp) { nonseekable_open(inode, filp); return 0; } static int riowd_release(struct inode *inode, struct file *filp) { return 0; } static long riowd_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) { static const struct watchdog_info info = { .options = WDIOF_SETTIMEOUT, .firmware_version = 1, .identity = DRIVER_NAME, }; void __user *argp = (void __user *)arg; struct riowd *p = riowd_device; unsigned int options; int new_margin; switch (cmd) { case WDIOC_GETSUPPORT: if (copy_to_user(argp, &info, sizeof(info))) return -EFAULT; break; case WDIOC_GETSTATUS: case WDIOC_GETBOOTSTATUS: if (put_user(0, (int __user *)argp)) return -EFAULT; break; case WDIOC_KEEPALIVE: riowd_writereg(p, riowd_timeout, WDTO_INDEX); break; case WDIOC_SETOPTIONS: if (copy_from_user(&options, argp, sizeof(options))) return -EFAULT; if (options & WDIOS_DISABLECARD) riowd_writereg(p, 0, WDTO_INDEX); else if (options & WDIOS_ENABLECARD) riowd_writereg(p, riowd_timeout, WDTO_INDEX); else return -EINVAL; break; case WDIOC_SETTIMEOUT: if (get_user(new_margin, (int __user *)argp)) return -EFAULT; if ((new_margin < 60) || (new_margin > (255 * 60))) return -EINVAL; riowd_timeout = (new_margin + 59) / 60; riowd_writereg(p, riowd_timeout, WDTO_INDEX); /* Fall */ case WDIOC_GETTIMEOUT: return put_user(riowd_timeout * 60, (int __user *)argp); default: return -EINVAL; }; return 0; } static ssize_t riowd_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos) { struct riowd *p = riowd_device; if (count) { riowd_writereg(p, riowd_timeout, WDTO_INDEX); return 1; } return 0; } static const struct file_operations riowd_fops = { .owner = THIS_MODULE, .llseek = no_llseek, .unlocked_ioctl = riowd_ioctl, .open = riowd_open, .write = riowd_write, .release = riowd_release, }; static struct miscdevice riowd_miscdev = { .minor = WATCHDOG_MINOR, .name = "watchdog", .fops = &riowd_fops }; static int riowd_probe(struct platform_device *op) { struct riowd *p; int err = -EINVAL; if (riowd_device) goto out; err = -ENOMEM; p = devm_kzalloc(&op->dev, sizeof(*p), GFP_KERNEL); if (!p) goto out; spin_lock_init(&p->lock); p->regs = of_ioremap(&op->resource[0], 0, 2, DRIVER_NAME); if (!p->regs) { pr_err("Cannot map registers\n"); goto out; } /* Make miscdev useable right away */ riowd_device = p; err = misc_register(&riowd_miscdev); if (err) { pr_err("Cannot register watchdog misc device\n"); goto out_iounmap; } pr_info("Hardware watchdog [%i minutes], regs at %p\n", riowd_timeout, p->regs); platform_set_drvdata(op, p); return 0; out_iounmap: riowd_device = NULL; of_iounmap(&op->resource[0], p->regs, 2); out: return err; } static int riowd_remove(struct platform_device *op) { struct riowd *p = platform_get_drvdata(op); misc_deregister(&riowd_miscdev); of_iounmap(&op->resource[0], p->regs, 2); return 0; } static const struct of_device_id riowd_match[] = { { .name = "pmc", }, {}, }; MODULE_DEVICE_TABLE(of, riowd_match); static struct platform_driver riowd_driver = { .driver = { .name = DRIVER_NAME, .of_match_table = riowd_match, }, .probe = riowd_probe, .remove = riowd_remove, }; module_platform_driver(riowd_driver); tools/perf/Documentation/tips.txt?id=0becc0ae5b42828785b589f686725ff5bc3b9b25'>patch) treebe6d0e1f37c38ed0a7dd5da2d4b1e93f0fb43101 /tools/perf/Documentation/tips.txt parent24c2503255d35c269b67162c397a1a1c1e02f6ce (diff)
x86/mce: Make timer handling more robust
Erik reported that on a preproduction hardware a CMCI storm triggers the BUG_ON in add_timer_on(). The reason is that the per CPU MCE timer is started by the CMCI logic before the MCE CPU hotplug callback starts the timer with add_timer_on(). So the timer is already queued which triggers the BUG. Using add_timer_on() is pretty pointless in this code because the timer is strictlty per CPU, initialized as pinned and all operations which arm the timer happen on the CPU to which the timer belongs. Simplify the whole machinery by using mod_timer() instead of add_timer_on() which avoids the problem because mod_timer() can handle already queued timers. Use __start_timer() everywhere so the earliest armed expiry time is preserved. Reported-by: Erik Veijola <erik.veijola@intel.com> Tested-by: Borislav Petkov <bp@alien8.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Borislav Petkov <bp@alien8.de> Cc: Tony Luck <tony.luck@intel.com> Link: http://lkml.kernel.org/r/alpine.DEB.2.20.1701310936080.3457@nanos Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Diffstat (limited to 'tools/perf/Documentation/tips.txt')