Skip to contentSkip to conversation

x86: Make XDBC work

PATCH In review Peter Zijlstra 3 participants 6 messages Mar 4, 2022 – Apr 20, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover x86: Make XDBC work

Hi!

These are the patches I needed to make XDBC go on my tigerlake NUC.

#1 lore raw

PATCH 1/3 x86/tsc: Be consistent about use_tsc_delay()

Currently loops_per_jiffy is set in tsc_early_init(), but then don't switch to delay_tsc, with the result that delay_loop is used with loops_per_jiffy set for delay_tsc.

Then in (late) tsc_init() lpj_fine is set (which is mostly unused) and after which use_tsc_delay() is finally called.

Move both loops_per_jiffy and use_tsc_delay() into tsc_enable_sched_clock() which is called the moment tsc_khz is determined, be it early or late. Keeping the lot consistent.

---
 arch/x86/kernel/tsc.c |    6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
arch/x86/kernel/tsc.c +3 -3
@@ -1487,6 +1487,9 @@ static unsigned long __init get_loops_pe
 
 static void __init tsc_enable_sched_clock(void)
 {
+	loops_per_jiffy = get_loops_per_jiffy();
+	use_tsc_delay();
+
 	/* Sanitize TSC ADJUST before cyc2ns gets initialized */
 	tsc_store_and_check_tsc_adjust(true);
 	cyc2ns_init_boot_cpu();
@@ -1502,8 +1505,6 @@ void __init tsc_early_init(void)
 		return;
 	if (!determine_cpu_tsc_frequencies(true))
 		return;
-	loops_per_jiffy = get_loops_per_jiffy();
-
 	tsc_enable_sched_clock();
 }
 
@@ -1537,7 +1538,6 @@ void __init tsc_init(void)
 		enable_sched_clock_irqtime();
 
 	lpj_fine = get_loops_per_jiffy();
-	use_tsc_delay();
 
 	check_system_tsc_reliable();
 
#2 lore raw

PATCH 2/3 usb: early: xhci-dbc: Remove duplicate keep parsing

In reply to Peter Zijlstra

The generic earlyprintk= parsing already parses the optional ",keep", no need to duplicate that in the xdbc driver.

---
 arch/x86/kernel/early_printk.c |    2 +-
 drivers/usb/early/xhci-dbc.c   |    5 ++---
 include/linux/usb/xhci-dbgp.h  |    2 +-
 3 files changed, 4 insertions(+), 5 deletions(-)
arch/x86/kernel/early_printk.c +1 -1
@@ -387,7 +387,7 @@ static int __init setup_early_printk(cha
 #endif
 #ifdef CONFIG_EARLY_PRINTK_USB_XDBC
 		if (!strncmp(buf, "xdbc", 4))
-			early_xdbc_parse_parameter(buf + 4);
+			early_xdbc_parse_parameter(buf + 4, keep);
 #endif
 
 		buf++;
drivers/usb/early/xhci-dbc.c +2 -3
@@ -599,7 +599,7 @@ static int __init xdbc_early_setup(void)
 	return 0;
 }
 
-int __init early_xdbc_parse_parameter(char *s)
+int __init early_xdbc_parse_parameter(char *s, int keep_early)
 {
 	unsigned long dbgp_num = 0;
 	u32 bus, dev, func, offset;
@@ -608,8 +608,7 @@ int __init early_xdbc_parse_parameter(ch
 	if (!early_pci_allowed())
 		return -EPERM;
 
-	if (strstr(s, "keep"))
-		early_console_keep = true;
+	early_console_keep = keep_early;
 
 	if (xdbc.xdbc_reg)
 		return 0;
include/linux/usb/xhci-dbgp.h +1 -1
@@ -15,7 +15,7 @@
 #define __LINUX_XHCI_DBGP_H
 
 #ifdef CONFIG_EARLY_PRINTK_USB_XDBC
-int __init early_xdbc_parse_parameter(char *s);
+int __init early_xdbc_parse_parameter(char *s, int keep_early);
 int __init early_xdbc_setup_hardware(void);
 void __init early_xdbc_register_console(void);
 #else
#3 lore raw

PATCH 3/3 usb: early: xhci-dbc: Fix xdbc number parsing

In reply to Peter Zijlstra

kstrtoul() assumes the string contains the number only and is \0 terminated, this is not the case, as such things like:

earlyprintk=xdbc1,keep

go completely sideways. Use simple_strtoul() instead.

---
 drivers/usb/early/xhci-dbc.c |    8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/usb/early/xhci-dbc.c +6 -2
@@ -603,6 +603,7 @@ int __init early_xdbc_parse_parameter(ch
 {
 	unsigned long dbgp_num = 0;
 	u32 bus, dev, func, offset;
+	char *e;
 	int ret;
 
 	if (!early_pci_allowed())
@@ -613,8 +614,11 @@ int __init early_xdbc_parse_parameter(ch
 	if (xdbc.xdbc_reg)
 		return 0;
 
-	if (*s && kstrtoul(s, 0, &dbgp_num))
-		dbgp_num = 0;
+	if (*s) {
+	       dbgp_num = simple_strtoul(s, &e, 10);
+	       if (s == e)
+		       dbgp_num = 0;
+	}
 
 	pr_notice("dbgp_num: %lu\n", dbgp_num);
 
Acked-by on v1 3/3
#4 lore raw

Re: [PATCH 3/3] usb: early: xhci-dbc: Fix xdbc number parsing

Quoted 23 lines · Peter Zijlstra

kstrtoul() assumes the string contains the number only and is \0 terminated, this is not the case, as such things like:

earlyprintk=xdbc1,keep

go completely sideways. Use simple_strtoul() instead.

---
 drivers/usb/early/xhci-dbc.c |    8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
drivers/usb/early/xhci-dbc.c +1 -0
@@ -603,6 +603,7 @@ int __init early_xdbc_parse_parameter(ch
 {
 	unsigned long dbgp_num = 0;
 	u32 bus, dev, func, offset;
+	char *e;
 	int ret;
 
 	if (!early_pci_allowed())

drivers/usb/early/xhci-dbc.c int __init early_xdbc_parse_parameter(ch · line 617

Quoted 6 lines
drivers/usb/early/xhci-dbc.c +1 -2
@@ -613,8 +614,11 @@ int __init early_xdbc_parse_parameter(ch
 	if (xdbc.xdbc_reg)
 		return 0;
 
-	if (*s && kstrtoul(s, 0, &dbgp_num))
-		dbgp_num = 0;
+	if (*s) {

Do you think we need a code comment here stating something like
"use deprecated simple_strtoul() as kstrtoul can't handle characters after the number"

drivers/usb/early/xhci-dbc.c int __init early_xdbc_parse_parameter(ch · line 622

Quoted 5 lines
drivers/usb/early/xhci-dbc.c +4 -0
+	       dbgp_num = simple_strtoul(s, &e, 10);
+	       if (s == e)
+		       dbgp_num = 0;
+	}
 

If not then PATCH 2/3 and 3/3 looks good to me
I don't know much about PATCH 1/3, the tsc change, but it didn't cause any issues for me either.

Thanks
-Mathias

#5 lore raw
In reply to Peter Zijlstra

Hi Peter,

Quoted 4 lines · Peter Zijlstra

Hi!

These are the patches I needed to make XDBC go on my tigerlake NUC.

I finally tested these patches (in mainline).
It's working well for me. Thanks.

Signature · 1 line
~Randy