| Message ID | 20090911041841.GA16844@rafazurita.homelinux.net (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
Hi, On Fri, Sep 11, 2009 at 01:18:41AM -0300, Rafael Ignacio Zurita wrote: > On Thu, Jul 30, 2009 at 02:25:26PM +0900, Paul Mundt wrote: > > On Wed, Jul 29, 2009 at 10:11:45PM -0300, Rafael Ignacio Zurita wrote: > > > Node 0: start_pfn = 0xd000, low = 0xe000 > > > Zone PFN ranges: > > > Normal 0x0000d000 -> 0x0000e000 > > > Movable zone start PFN for each node > > > early_node_map[1] active PFN ranges > > > 0: 0x0000d000 -> 0x0000e000 > > > Built 1 zonelists in Zone order, mobility grouping off. Total pages: 4064 > > > Kernel command line: mem=16M init=/bin/sh earlyprintk=serial > > > NR_IRQS:256 > > > PID hash table entries: 64 (order: 6, 256 bytes) > > > Console: colour dummy device 80x25 > > > console handover: boot [sercon0] -> real [tty0] > > > > > > and I guess that the system hangs. > > > Should I try other printks in other places? Suggestions? > > > > > earlyprintk=serial,keep > > Thanks, I get more output for debug. > Now I got the below kernel messages (sorry for the long output): > ... > > The important message (I guess) is : > > 2 locks held by swapper/0: > #0: (clock_list_sem){......}, at: [<8d009928>] clk_register+0x58/0xf0 > #1: (clock_list_sem){......}, at: [<8d009c6c>] clk_get_sys+0x1c/0xd0 > Unable to handle kernel NULL pointer dereference at virtual address 0000001c > > > So I read for a while arch/sh/kernel/cpu/clock.c > trying to understand if I see some problem. Both, int clk_register(struct clk *clk) > and struct clk *clk_get_sys(const char *dev_id, const char *con_id) > do mutex_lock(&clock_list_sem) and mutex_unlock(&clock_list_sem). > > Then, I did this crazy change below: > > diff -upr linux-2.6.orig/arch/sh/kernel/cpu/clock.c linux-2.6.new/arch/sh/kernel/cpu/clock.c > --- linux-2.6.orig/arch/sh/kernel/cpu/clock.c 2009-09-11 01:09:35.000000000 -0300 > +++ linux-2.6.new/arch/sh/kernel/cpu/clock.c 2009-09-11 01:12:01.000000000 -0300 > @@ -266,7 +266,7 @@ int clk_register(struct clk *clk) > if (clk->node.next || clk->node.prev) > return 0; > > - mutex_lock(&clock_list_sem); > + /* mutex_lock(&clock_list_sem); */ > > INIT_LIST_HEAD(&clk->children); > clk->usecount = 0; > @@ -279,7 +279,7 @@ int clk_register(struct clk *clk) > list_add(&clk->node, &clock_list); > if (clk->ops && clk->ops->init) > clk->ops->init(clk); > - mutex_unlock(&clock_list_sem); > + /* mutex_unlock(&clock_list_sem); */ > > return 0; > } > > > and after building the kernel boots and goes further: I did printks and I tried to do a tracing : From arch/sh/kernel/cpu/clock-cpg.c : static struct clk *onchip_clocks[] = { &master_clk, &peripheral_clk, &bus_clk, &cpu_clk, }; int __init __deprecated cpg_clk_init(void) { int i, ret = 0; for (i = 0; i < ARRAY_SIZE(onchip_clocks); i++) { struct clk *clk = onchip_clocks[i]; arch_init_clk_ops(&clk->ops, i); if (clk->ops) ret |= clk_register(clk); } return ret; } When i==3, clk_register() for cpu_clk looks like a breakage. clk_register() from arch/sh/kernel/cpu/clock.c says : int clk_register(struct clk *clk) { if (clk == NULL || IS_ERR(clk)) return -EINVAL; /* * trap out already registered clocks */ if (clk->node.next || clk->node.prev) return 0; mutex_lock(&clock_list_sem); INIT_LIST_HEAD(&clk->children); clk->usecount = 0; if (clk->parent) list_add(&clk->sibling, &clk->parent->children); else list_add(&clk->sibling, &root_clks); list_add(&clk->node, &clock_list); if (clk->ops && clk->ops->init) clk->ops->init(clk); mutex_unlock(&clock_list_sem); return 0; } EXPORT_SYMBOL_GPL(clk_register); There is a first mutex_lock, and then, it calls clk->ops->init(clk). From arch/sh/kernel/cpu/sh3/clock-sh7709.c : static void set_bus_parent(struct clk *clk) { struct clk *bus_clk = clk_get(NULL, "bus_clk"); clk->parent = bus_clk; clk_put(bus_clk); } static struct clk_ops sh7709_cpu_clk_ops = { .init = set_bus_parent, .recalc = cpu_clk_recalc, }; static struct clk_ops *sh7709_clk_ops[] = { &sh7709_master_clk_ops, &sh7709_module_clk_ops, &sh7709_bus_clk_ops, &sh7709_cpu_clk_ops, }; set_bus_parent() calls clk_get(). Then from arch/sh/kernel/cpu/clock.c : struct clk *clk_get(struct device *dev, const char *id) { const char *dev_id = dev ? dev_name(dev) : NULL; struct clk *p, *clk = ERR_PTR(-ENOENT); int idno; clk = clk_get_sys(dev_id, id); ... struct clk *clk_get_sys(const char *dev_id, const char *con_id) { struct clk *clk; mutex_lock(&clock_list_sem); clk = clk_find(dev_id, con_id); mutex_unlock(&clock_list_sem); return clk ? clk : ERR_PTR(-ENOENT); } EXPORT_SYMBOL_GPL(clk_get_sys); Again, a new mutex_lock. Is that correct? I mean, printks stop with the last for (i = 0; i < ARRAY_SIZE(onchip_clocks); i++) { in int __init __deprecated cpg_clk_init(void) So it looks for me like two locks before the first one unlocks. If I am reading well, that should not be a proper situation right? There I would like to see the two below lines if (clk->ops && clk->ops->init) clk->ops->init(clk); out from lock-unlock section. Thoughts? Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-sh" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff -upr linux-2.6.orig/arch/sh/kernel/cpu/clock.c linux-2.6.new/arch/sh/kernel/cpu/clock.c --- linux-2.6.orig/arch/sh/kernel/cpu/clock.c 2009-09-11 01:09:35.000000000 -0300 +++ linux-2.6.new/arch/sh/kernel/cpu/clock.c 2009-09-11 01:12:01.000000000 -0300 @@ -266,7 +266,7 @@ int clk_register(struct clk *clk) if (clk->node.next || clk->node.prev) return 0; - mutex_lock(&clock_list_sem); + /* mutex_lock(&clock_list_sem); */ INIT_LIST_HEAD(&clk->children); clk->usecount = 0; @@ -279,7 +279,7 @@ int clk_register(struct clk *clk) list_add(&clk->node, &clock_list); if (clk->ops && clk->ops->init) clk->ops->init(clk); - mutex_unlock(&clock_list_sem); + /* mutex_unlock(&clock_list_sem); */ return 0; }