Linux-2.1.109.. preliminary code freeze.

Ok, it's out there now in all its glory...
2.1.109 does the following thing:

 - CPU detection in C code (and thus much easier to expand upon,
   especially as it's all thrown away after booting now that it is
   "initfunc()").  This should finally get the Cyrix case right, for
   example. Please test.
 - too meny people convinced me that sendfile() really wants to act like
   writep().
 - sound driver updates from Alan.
 - console updates, so now we have the full old functionality again as far
   as I'm concerned (but I'm sure people will tell me something is still
   missing)
 - task switch and user space return cleanly handles bad segment
   descriptors etc, so people shoul dno longer be able to cause kernel
   messages by misusing the LDT (and I was just informed that you could
   actually completely hang a 2.1.x SMP kernel by doing nasty things -
   this fixes it)
 - wine should work again thanks to Bill Hawes (other LDT fixes)
 - de4x5 driver update
 - token ring driver update
 - ppp driver update
 - coda-fs update
 - "shared writable" bug fixed (thanks to a lot of people for testing and
   working on this - the actual fix was trivial once the problem was
   understood)

In addition, I've spent a large part of my day running with a 12MB
machine, and low-memory behaviour seems to be reasonable. People who have
been unhappy with low-memory behaviour should check out 109 and comment on
it - the heuristics are fairly different, and seem to be better.

As of this release, I won't be looking at the "incoming" directory at the
linux-patches site any more. I'll only be looking at "urgent" things, on
the theory that I'm (a) lazy and (b) getting into code freeze.

If you have important patches in "incoming", feel free to move them to
"urgent". However, I will warn that if I don't consider them to be 2.2
material, I'll just move them to "discarded".

The same goes for patches in email. I will accept patches, but I've just
raised the bar for acception.

                Linus

CREDITS

@@ -125,6 +125,16 @@ S: 633 South 550 East
 S: Provo, Utah 84606
 S: USA
 
+M: Krzysztof G. Baranowski
+E: kgb@manjak.knm.org.pl
+P: 1024/FA6F16D1 96 D1 1A CF 5F CA 69 EC  F9 4F 36 1F 6D 60 7B DA
+D: Maintainer of the System V file system.
+D: SystemV fs update for 2.1.x dcache.
+D: Various bugfixes.
+S: ul. Koscielna 12a
+S: 62-300 Wrzesnia
+S: Poland
+
 N: Peter Bauer
 E: 100136.3530@compuserve.com
 D: Driver for depca-ethernet-board

Documentation/Configure.help

@@ -7556,24 +7556,6 @@ CONFIG_MTRR
 
   See Documentation/mtrr.txt for more information.
 
-Cyrix low power feature
-CONFIG_CYRIX_SUSP
-  Cyrix 6x86 family CPUs include a feature named "Suspend on Halt".
-  When enabled, a single "halt" instruction puts the CPU in suspend mode
-  till the next interruption. In suspend mode, the CPU dissipates less
-  than the fiftieth of its nominal power ; the good news is that Linux
-  executes a halt instruction when it has nothing to, so that your CPU
-  will spend its idle time in suspend mode. Note that you can also
-  manually activate/deactivate this feature by using the set6x86 tool
-  available at :
-      ftp://tsx-11.mit.edu/pub/linux/sources/sbin/set6x86.tar.gz
-
-  There's no risk to use this. The only reason why you would say no is
-  if you are enough of hearing you fans changing speed because of the
-  difference in power consumption between idle and work time.
-
-  Please email any comments to : tarreau@aemiaif.lip6.fr
-
 Main CPU frequency, only for DEC alpha machine
 CONFIG_FT_ALPHA_CLOCK
   On some DEC Alpha machines the CPU clock frequency cannot be

Documentation/sound/AWE32

@@ -54,8 +54,7 @@ but the wavetable needs THREE! My working string is:
 (IO 3 (BASE 0x0E20)) (ACT Y) ))"
 
 Resources 0x0620, 0x0A20 and 0x0E20 should work. Other on-board devices: 
-Gameport and StereoEnhance are not needed. In my case StereoEnhance didn't
-even work!
+Gameport and StereoEnhance are not required to be inited.
 
 Now you can execute "isapnp /etc/isapnp.conf". No errors should be reported.
 If you correctly installed isapnptools, then isapnp will run every boot time.
@@ -64,22 +63,20 @@ If you correctly installed isapnptools, then isapnp will run every boot time.
 because the AWE32 driver is included in them.  ATTENTION!  In kernels 2.1.102, 
 2.1.103, 2.1.104-pre1 and 2.1.104 (not the others) the lowlevel sound driver
 is not working. You should use the patch available at 
-http://members.xoom.com/yar. If you are using stable kernel releases 
-2.0.x (I am not sure about 2.0.34), then get the latest version (3.8s9) of 
+http://members.xoom.com/yar/history.html. If you are using stable kernel
+releases 2.0.x, then get the latest version (3.8s9) of 
 OSS/Free at http://members.xoom.com/yar/ossfree38s9-linux20x.tar.gz
-and untar it in /usr/src/ (assuming you keep your kernel source in
-/usr/src/linux).  Then go to /usr/src/linux/ and view
-the README file. That file contains info about kernel compilation and
-installation. If you use kernel version 2.0.x or version <= 2.1.104 go 
-to substep a, on 2.1.105 or later go to step b.
+and gunzip/untar it in /usr/src/ (assuming you keep your kernel source in
+/usr/src/linux).  Then go to /usr/src/linux/ and view the README file. That
+file contains info about kernel compilation and installation.
 
 In "make (x,menu)config" select in "Sound":
 "Sound card support", "100% Sound Blaster compatibles (SB16/32/64, ESS, Jazz16) support",
 "Generic OPL2/OPL3 FM synthesizer support" and "FM synthesizer (YM3812/OPL-3) 
 support" as <M> (module).
 
-If you use kernel version 2.0.x or version <= 2.1.104 skip substep a, on 
-2.1.105 or later go through it.
+If you use kernel version 2.0.x or version 2.1.y (y <= 2.1.104) skip substep a,
+on 2.1.105 or later go through it.
 
 substep a:
 In "make (x,menu)config" select in "Sound":
@@ -90,7 +87,7 @@ In "Additional low level sound drivers":
 Select "Additional low level sound drivers" as [y] (or [*] (yes)) (If it is not
 available as [y], select it as <M> (module))
 Now recompile the kernel (make dep; make (b)zImage; make modules; 
-make modules_install) and boot new kernel.
+make modules_install), update your boot loader and boot new kernel.
 
 6) Now download awesfx program 
 http://members.xoom.com/yar/awesfx-0.4.2.tgz here. Compile it.

arch/i386/kernel/head.S

@@ -4,8 +4,7 @@
  *  Copyright (C) 1991, 1992  Linus Torvalds
  *
  *  Enhanced CPU detection and feature setting code by Mike Jagdis
- *  and Martin Mares, November 1997, and partially recoded and moved
- *  to setup.c by Willy Tarreau, July 1998.
+ *  and Martin Mares, November 1997.
  */
 
 .text
@@ -22,9 +21,7 @@
 #define CL_OFFSET	0x90022
 
 /*
- * References to members of the boot_cpu_data structure. Member
- * definitions other than X86 and X86_HARD_MATH may be removed
- * since they're no more used here.
+ * References to members of the boot_cpu_data structure.
  */
 
 #define CPU_PARAMS	SYMBOL_NAME(boot_cpu_data)
@@ -141,13 +138,17 @@ startup_32:
 #ifdef __SMP__
 checkCPUtype:
 #endif
+
+	movl $-1,X86_CPUID		#  -1 for no CPUID initially
+
+/* check if it is 486 or 386. */
 /*
- * check if the cpu is >=486 or <=386.
  * XXX - this does a lot of unnecessary setup.  Alignment checks don't
  * apply at our cpl of 0 and the stack ought to be aligned already, and
  * we don't need to preserve eflags.
  */
 
+	movl $3,X86		# at least 386
 	pushfl			# push EFLAGS
 	popl %eax		# get EFLAGS
 	movl %eax,%ecx		# save original EFLAGS
@@ -157,21 +158,56 @@ checkCPUtype:
 	pushfl			# get new EFLAGS
 	popl %eax		# put it in eax
 	xorl %ecx,%eax		# change in flags
-	pushl %ecx		# get original EFLAGS
-	popfl			# restore EFLAGS
 	andl $0x40000,%eax	# check if AC bit changed
-	movl %cr0,%eax		# prepares new EAX
-	je 1f			# at most a 386
+	je is386
 
 	movl $4,X86		# at least 486
+	movl %ecx,%eax
+	xorl $0x200000,%eax	# check ID flag
+	pushl %eax
+	popfl			# if we are on a straight 486DX, SX, or
+	pushfl			# 487SX we can't change it
+	popl %eax
+	xorl %ecx,%eax
+	pushl %ecx		# restore original EFLAGS
+	popfl
+	andl $0x200000,%eax
+	je is486
+
+	/* get vendor info */
+	xorl %eax,%eax			# call CPUID with 0 -> return vendor ID
+	cpuid
+	movl %eax,X86_CPUID		# save CPUID level
+	movl %ebx,X86_VENDOR_ID		# lo 4 chars
+	movl %edx,X86_VENDOR_ID+4	# next 4 chars
+	movl %ecx,X86_VENDOR_ID+8	# last 4 chars
+
+	orl %eax,%eax			# do we have processor info as well?
+	je is486
+
+	movl $1,%eax		# Use the CPUID instruction to get CPU type
+	cpuid
+	movb %al,%cl		# save reg for future use
+	andb $0x0f,%ah		# mask processor family
+	movb %ah,X86
+	andb $0xf0,%al		# mask model
+	shrb $4,%al
+	movb %al,X86_MODEL
+	andb $0x0f,%cl		# mask mask revision
+	movb %cl,X86_MASK
+	movl %edx,X86_CAPABILITY
+
+is486:
+	movl %cr0,%eax		# 486 or better
 	andl $0x80000011,%eax	# Save PG,PE,ET
 	orl $0x50022,%eax	# set AM, WP, NE and MP
 	jmp 2f
-1:
-	movl $3,X86		# at most 386
+
+is386:	pushl %ecx		# restore original EFLAGS
+	popfl
+	movl %cr0,%eax		# 386
 	andl $0x80000011,%eax	# Save PG,PE,ET
 	orl $2,%eax		# set MP
-
 2:	movl %eax,%cr0
 	call check_x87
 #ifdef __SMP__
@@ -209,7 +245,6 @@ checkCPUtype:
 	pushl %eax
 	pushl %eax
 	cld			# gcc2 wants the direction flag cleared at all times
-	call SYMBOL_NAME(refine_cpu_detection)	# get more info about cpu
 	call SYMBOL_NAME(start_kernel)
 L6:
 	jmp L6			# main should never return here, but

arch/i386/kernel/process.c

@@ -423,6 +423,54 @@ void show_regs(struct pt_regs * regs)
 		0xffff & regs->xds,0xffff & regs->xes);
 }
 
+/*
+ * Allocation and freeing of basic task resources.
+ *
+ * NOTE! The task struct and the stack go together
+ *
+ * The task structure is a two-page thing, and as such
+ * not reliable to allocate using the basic page alloc
+ * functions. We have a small cache of structures for
+ * when the allocations fail..
+ *
+ * This extra buffer essentially acts to make for less
+ * "jitter" in the allocations..
+ */
+#define EXTRA_TASK_STRUCT	16
+static struct task_struct * task_struct_stack[EXTRA_TASK_STRUCT];
+static int task_struct_stack_ptr = -1;
+
+struct task_struct * alloc_task_struct(void)
+{
+	int index;
+	struct task_struct *ret;
+
+	index = task_struct_stack_ptr;
+	if (index >= EXTRA_TASK_STRUCT/2)
+		goto use_cache;
+	ret = (struct task_struct *) __get_free_pages(GFP_KERNEL,1);
+	if (!ret) {
+		index = task_struct_stack_ptr;
+		if (index >= 0) {
+use_cache:
+			ret = task_struct_stack[index];
+			task_struct_stack_ptr = index-1;
+		}
+	}
+	return ret;
+}
+
+void free_task_struct(struct task_struct *p)
+{
+	int index = task_struct_stack_ptr+1;
+
+	if (index < EXTRA_TASK_STRUCT) {
+		task_struct_stack[index] = p;
+		task_struct_stack_ptr = index;
+	} else
+		free_pages((unsigned long) p, 1);
+}
+
 void release_segments(struct mm_struct *mm)
 {
 	void * ldt = mm->segments;

arch/i386/kernel/setup.c

@@ -3,9 +3,8 @@
  *
  *  Copyright (C) 1995  Linus Torvalds
  *
- *  Enhanced CPU detection and feature setting code by Mike Jagdis
- *  and Martin Mares, November 1997, and partially recoded and moved
- *  form head.S by Willy Tarreau, July 1998.
+ *  Enhanced CPU type detection by Mike Jagdis, Patrick St. Jean
+ *  and Martin Mares, November 1997.
  */
 
 /*
@@ -45,6 +44,7 @@
 
 char ignore_irq13 = 0;		/* set if exception 16 works */
 struct cpuinfo_x86 boot_cpu_data = { 0, 0, 0, 0, -1, 1, 0, 0, -1 };
+static char Cx86_step[8];       /* decoded Cyrix step number */
 
 /*
  * Bus types ..
@@ -229,312 +229,170 @@ __initfunc(void setup_arch(char **cmdline_p,
 #endif
 }
 
-/*
- *	Detection of CPU model.
- */
-
-extern inline void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx)
-{
-	__asm__("cpuid"
-		: "=a" (*eax),
-		  "=b" (*ebx),
-		  "=c" (*ecx),
-		  "=d" (*edx)
-		: "a" (op)
-		: "cc");
-}
-
-/*
- *	Try to enable ID flag in EFLAGS, in order to allow CPUID instruction.
- *	Return 0 if not possible, anything else if it succeeded.
- *	
- */
-
-extern __u32 enable_cpuid(void)
-{
-	__u32 ret;
-	__asm__ __volatile__ (
-		"pushfl\n\t"
-		"popl %%eax\n\t"
-		"movl %%eax,%%ecx\n\t"
-		"xorl $0x200000,%%eax\n\t"
-		"pushl %%eax\n\t"
-		"popfl\n\t"
-		"pushfl\n\t"
-		"popl %%eax\n\t"
-		"xorl %%ecx,%%eax\n\t"
-		"andl $0x200000,%%eax\n\t" /* get only the ID flag */
-		"orl  %%eax,%%ecx\n\t" /* and force it into EFLAGS */
-		"pushl %%ecx\n\t"
-		"popfl\n\t"
-		: "=a" (ret) : : "cc", "%ecx");
-	return ret;
-}
-
-/*
- *	Returns 1 only if the DIV instruction keeps flags. This is the case on
- *	all Cyrix, and on Intel's PPro/PII, but they can be distinguished from
- *	a Cyrix. In other cases, it returns 0.
- */
-
-extern inline __u32 div_keeps_flags(void)
-{
-	__u32 ret;
-	__asm__ __volatile__ (
-		"xorl %%eax,%%eax\n\t"
-		"sahf\n\t"
-		"movb $5,%%al\n\t"
-		"movb $2,%%bl\n\t"
-		"divb %%bl\n\t"
-		"lahf\n\t"
-		"cmpb $2,%%ah\n\t"
-		"setzb %%al\n\t"
-		"movzbl %%al, %%eax\n\t"	      
-		: "=a" (ret)
-		:
-		: "cc", "%ebx");
-	return ret;
-}
-
-/*
- * Functions for getting/setting Cyrix's control registers (CCR).
- */
-__initfunc(static inline void setCx86(__u8 reg, __u8 val))
-{
-	__asm__ __volatile__ ("outb %%al,$0x22\n\t"
-			      "mov %1,%%al\n\t"
-			      "outb %%al,$0x23\n\t"
-			      :
-			      : "a" (reg), "q" (val)
-			      : "eax");
-}
-
-__initfunc(static inline __u8 getCx86(__u8 reg))
-{
-	__u8 ret;
-	__asm__ __volatile__ ("outb %%al,$0x22\n\t"
-			      "inb $0x23,%%al\n\t"
-			      : "=a" (ret)
-			      : "a" (reg));
-
-	return ret;
-}
-
-__initfunc(static void refine_cyrix_detection(struct cpuinfo_x86 *c))
+__initfunc(static int amd_model(struct cpuinfo_x86 *c))
 {
-	__u8 ccr3;
-
-	ccr3 = getCx86(0xc3);
-	setCx86(0xc3, ccr3 | 0x80);
-	getCx86(0xc0);			/* dummy read to change the bus */
-	if (getCx86(0xc3) == ccr3)	/* ccr3 unchanged : DEVID not supported */
-		return;			/* let's end here. */
-
-	setCx86(0xc3, ccr3);	/* restore ccr3 */
+	unsigned int n, dummy, *v;
 
-	/*
-	 * Now, fill x86_mask/model with DEVID.
+	/* Actually we must have cpuid or we could never have
+	 * figured out that this was AMD from the vendor info :-).
 	 */
-	c->x86_model = getCx86(0xfe);	/* DIR0 to model */
-	c->x86_mask  = getCx86(0xff);	/* DIR1 to mask */
-
-	if ((c->x86_model & 0xF0) == 0x30) {	/* 6x86(L)! we got it !*/
-		/*
-		 * we'll do here a few optimizations.
-		 */
-
-		/* since 6x86(L) doesn't support SMP, we'll deactivate all LOCK# cycles
-		 * so that even if the kernel has been compiled with the -D__SMP__ option
-		 * as this is often the case, lock instructions won't hurt performance.
-		 */
-		setCx86(0xc1, getCx86(0xc1) | 0x10);	/* assert NO_LOCK */
-		setCx86(0xc3, (getCx86(0xc3) | 0x10) & 0x1F);	/* enables high registers */
-		setCx86(0xe8, getCx86(0xe8) | 0x10);	/* enable Directory Table Entry Cache */
-		setCx86(0xe9, getCx86(0xe9) | 0x01);	/* enable cache Write-Through Allocate */
-		setCx86(0xc3, getCx86(0xc3) & 0x0F);	/* disable high registers */
-	}
 
-#ifdef CONFIG_CYRIX_SUSP
-	/*
-	 * This option enables the low power feature of the 6x86: it can
-	 * suspend on halt without any performance hurt, as Cyrix claims.
-	 *
-	 * This option is not exported yet, as I wonder about the cycle
-	 * counter..			Linus
-	 */
-	if (((c->x86_model & 0xF0) == 0x30) ||	/* 6x86(L) */
-	    ((c->x86_model & 0xF0) == 0x50)) {	/* 6x86MX */
-		/*
-		 * suspend on halt : this amazingly reduces power consumption
-		 * and cpu overheat.
-		 */
-		setCx86(0xc2, getCx86(0xc2) | 0x08);
-	}
-#endif
+	cpuid(0x80000000, &n, &dummy, &dummy, &dummy);
+	if (n < 4)
+		return 0;
+	v = (unsigned int *) c->x86_model_id;
+	cpuid(0x80000002, &v[0], &v[1], &v[2], &v[3]);
+	cpuid(0x80000003, &v[4], &v[5], &v[6], &v[7]);
+	cpuid(0x80000004, &v[8], &v[9], &v[10], &v[11]);
+	c->x86_model_id[48] = 0;
+	return 1;
 }
 
 /*
- *	This function uses CPUID to get some precisions about the current
- *	cpu. It's called directly from head.S, just before branching to
- *	start_kernel. It could have been called from other points, but in
- *	any case, it must be called before setup_arch() because this function
- *	uses (and changes) x86_capability which, of course, must have been set.
+ * Use the Cyrix DEVID CPU registers if avail. to get more detailed info.
  */
-__initfunc(void refine_cpu_detection(void))
+__initfunc(static void do_cyrix_devid(struct cpuinfo_x86 *c))
 {
-	unsigned int dummy, r;
-	struct cpuinfo_x86 *c=&boot_cpu_data;
-
-	c->cpuid_level=-1;
-	if (enable_cpuid()) {
-	/*
-	 * CPUID is supported by this cpu. It's a recent one.
-	 * All Pentium, PentiumPro, PentiumII, AmdK5, AmdK6, 6x86MX and some
-	 * latests 486 will go directly here.
-	 */
-
-		/* get vendor_id string */
-		cpuid(0, &c->cpuid_level,
-		         (int *)&c->x86_vendor_id[0],
-		         (int *)&c->x86_vendor_id[8],
-		         (int *)&c->x86_vendor_id[4]);
-		/* get model and mask */
-		cpuid(1, &r, &dummy, &dummy, (int *)&c->x86_capability);
-		c->x86       = (r>>8) & 0xF;
-		c->x86_model = (r>>4) & 0xF;
-		c->x86_mask  = r & 0xF;
-	} else {
-		/*
-		 * CPUID doesn't seem to be supported. But if the cpu is a Cyrix, we
-		 * can try to activate it. If not, it's simply an old 386 or 486 from
-		 * which we won't get anything more.
-		 */
-
-		__u8 ccr3;
-
-		if (c->x86 <= 3)	/* nothing more can be done, it's a 386. */
-			return;
-
-		/*
-		 * The following test can distinguish a Cyrix from an other 486 cpu.
-		 * As Intel's PPro and PII behaves like cyrix, we must use this code
-		 * only when we are certain that the cpu *is not* a PPro/PII.
-		 */
-		if (!div_keeps_flags())
-			return;	/* just a simple 486. let's forget CPUID. */
-
-		/*
-		 * now we know we have a Cyrix. let's activate its CPUID if it has one.
-		 */
-		setCx86(0xC3, ((ccr3=getCx86(0xC3)) & 0x0F) | 0x10); /* enable all CCRs */
-		setCx86(0xE8, getCx86(0xE8) | 0x80);
-		setCx86(0xC3, ccr3);
-
-		/*
-		 * first we assume it's a cx486 because it doesn't have cpuid.
-		 */
-		*(long long int*)&c->x86_vendor_id = 0x7869727943ULL; /* "Cyrix\0\0\0" */
-		c->x86_model = 0xFE;	/* Cx486 */
-		c->x86_mask  = 0;
-
-		/*
-		 * if CPUID is supported now, we'll use it, but won't trust it.
-		 */
-		if (enable_cpuid()) {
-			cpuid(0, &c->cpuid_level,
-			      (int *)&c->x86_vendor_id[0],
-			      (int *)&c->x86_vendor_id[8],
-			      (int *)&c->x86_vendor_id[4]);
-			cpuid(1, &r, &dummy, &dummy, (int*)&c->x86_capability);
-			c->x86       = (r>>8) & 0xF; /* cpu family */
-			c->x86_model = (r>>4) & 0xF;
-			c->x86_mask  = r & 0xF;
+	unsigned char ccr2, ccr3;
+
+	/* we test for DEVID by checking whether CCR3 is writable */
+	cli();
+	ccr3 = getCx86(CX86_CCR3);
+	setCx86(CX86_CCR3, ccr3 ^ 0x80);
+	getCx86(0xc0);   /* dummy to change bus */
+
+	if (getCx86(CX86_CCR3) == ccr3) {       /* no DEVID regs. */
+		ccr2 = getCx86(CX86_CCR2);
+		setCx86(CX86_CCR2, ccr2 ^ 0x04);
+		getCx86(0xc0);  /* dummy */
+
+		if (getCx86(CX86_CCR2) == ccr2) /* old Cx486SLC/DLC */
+			c->x86_model = 0xfd;
+		else {                          /* Cx486S A step */
+			setCx86(CX86_CCR2, ccr2);
+			c->x86_model = 0xfe;
 		}
 	}
-		
-	/*
-	 * If we have any Cyrix, even the latest one, we'll read DEVID.
-	 * As it was assumed in head.S, we suppose that Cyrix's DEVID is
-	 * more accurate than its CPUID. So if available, we read it.
-	 */
-	if (*(int *)&c->x86_vendor_id == 0x69727943) {	/* "Cyri"... */
-		refine_cyrix_detection(c);
+	else {
+		setCx86(CX86_CCR3, ccr3);  /* restore CCR3 */
+
+		/* read DIR0 and DIR1 CPU registers */
+		c->x86_model = getCx86(CX86_DIR0);
+		c->x86_mask = getCx86(CX86_DIR1);
 	}
-	c->x86_vendor_id[12]=0; /* terminates the vendor id string. */
+	sti();
 }
 
+static char Cx86_model[][9] __initdata = {
+	"Cx486", "Cx486", "5x86 ", "6x86", "MediaGX ", "6x86MX ",
+	"M II ", "Unknown"
+};
+static char Cx486_name[][5] __initdata = {
+	"SLC", "DLC", "SLC2", "DLC2", "SRx", "DRx",
+	"SRx2", "DRx2"
+};
+static char Cx486S_name[][4] __initdata = {
+	"S", "S2", "Se", "S2e"
+};
+static char Cx486D_name[][4] __initdata = {
+	"DX", "DX2", "?", "?", "?", "DX4"
+};
+static char Cx86_cb[] __initdata = "?.5x Core/Bus Clock";
+static char cyrix_model_mult1[] __initdata = "12??43";
+static char cyrix_model_mult2[] __initdata = "12233445";
+static char cyrix_model_oldstep[] __initdata = "A step";
 
-__initfunc(static int cyrix_model(struct cpuinfo_x86 *c))
+__initfunc(static void cyrix_model(struct cpuinfo_x86 *c))
 {
-	int nr = c->x86_model;
+	unsigned char dir0_msn, dir0_lsn, dir1;
 	char *buf = c->x86_model_id;
+	const char *p = NULL;
 
-	/* Cyrix claims they have a TSC, but it is broken */
-	c->x86_capability &= ~X86_FEATURE_TSC;
+	do_cyrix_devid(c);
 
-	/* Note that some of the possibilities this decoding allows
-	 * have never actually been manufactured - but those that
-	 * do actually exist are correctly decoded.
-	 */
-	if (nr < 0x20) {
-		strcpy(buf, "Cx486");
-		if (!(nr & 0x10)) {
-			sprintf(buf+5, "%c%s%c",
-				(nr & 0x01) ? 'D' : 'S',
-				(nr & 0x04) ? "Rx" : "LC",
-				(nr & 0x02) ? '2' : '\000');
-		} else if (!(nr & 0x08)) {
-			sprintf(buf+5, "S%s%c",
-				(nr & 0x01) ? "2" : "",
-				(nr & 0x02) ? 'e' : '\000');
-		} else {
-			sprintf(buf+5, "DX%c",
-				nr == 0x1b ? '2'
-					: (nr == 0x1f ? '4' : '\000'));
-		}
-	} else if (nr >= 0x20 && nr <= 0x4f) {	/* 5x86, 6x86 or Gx86 */
-		char *s = "";
-		if (nr >= 0x30 && nr < 0x40) {	/* 6x86 */
-			if (c->x86 == 5 && (c->x86_capability & X86_FEATURE_CX8))
-				s = "L";	/* 6x86L */
-			else if (c->x86 == 6)
-				s = "MX";	/* 6x86MX */
-			}
-		sprintf(buf, "%cx86%s %cx Core/Bus Clock",
-			"??56G"[nr>>4],
-			s,
-			"12??43"[nr & 0x05]);
-	} else if (nr >= 0x50 && nr <= 0x5f) {	/* Cyrix 6x86MX */
-		sprintf(buf, "6x86MX %c%sx Core/Bus Clock",
-			"12233445"[nr & 0x07],
-			(nr && !(nr&1)) ? ".5" : "");
-	} else if (nr >= 0xfd && c->cpuid_level < 0) {
-		/* Probably 0xfd (Cx486[SD]LC with no ID register)
-		 * or 0xfe (Cx486 A step with no ID register).
-		 */
-		strcpy(buf, "Cx486");
-	} else
-		return 0;	/* Use CPUID if applicable */
-	return 1;
-}
+	dir0_msn = c->x86_model >> 4;
+	dir0_lsn = c->x86_model & 0xf;
+	dir1 = c->x86_mask;
 
-__initfunc(static int amd_model(struct cpuinfo_x86 *c))
-{
-	unsigned int n, dummy, *v;
+	/* common case stepping number -- exceptions handled below */
+	sprintf(Cx86_step, "%d.%d", (dir1 >> 4) + 1, dir1 & 0x0f);
 
-	/* Actually we must have cpuid or we could never have
-	 * figured out that this was AMD from the vendor info :-).
+	/* Now cook; the original recipe is by Channing Corn, from Cyrix.
+	 * We do the same thing for each generation: we work out
+	 * the model, multiplier and stepping.
 	 */
+	switch (dir0_msn) {
+		unsigned char tmp;
+
+	case 0: /* Cx486SLC/DLC/SRx/DRx */
+		p = Cx486_name[dir0_lsn & 7];
+		break;
+
+	case 1: /* Cx486S/DX/DX2/DX4 */
+		p = (dir0_lsn & 8) ? Cx486D_name[dir0_lsn & 5]
+			: Cx486S_name[dir0_lsn & 3];
+		break;
+
+	case 2: /* 5x86 */
+		Cx86_cb[2] = cyrix_model_mult1[dir0_lsn & 5];
+		p = Cx86_cb+2;
+		break;
+
+	case 3: /* 6x86/6x86L */
+		Cx86_cb[1] = ' ';
+		Cx86_cb[2] = cyrix_model_mult1[dir0_lsn & 5];
+		if (dir1 > 0x21) { /* 686L */
+			Cx86_cb[0] = 'L';
+			p = Cx86_cb;
+			Cx86_step[0]++;
+		} else             /* 686 */
+			p = Cx86_cb+1;
+		break;
+
+	case 4: /* MediaGX/GXm */
+		/* GXm supports extended cpuid levels 'ala' AMD */
+		if (c->cpuid_level == 2) {
+			amd_model(c);  /* get CPU marketing name */
+			return;
+		}
+		else {  /* MediaGX */
+			Cx86_cb[2] = (dir0_lsn & 1) ? '3' : '4';
+			p = Cx86_cb+2;
+			Cx86_step[0] = (dir1 & 0x20) ? '1' : '2';
+		}
+		break;
+
+        case 5: /* 6x86MX/M II */
+		/* the TSC is broken (for now) */
+		c->x86_capability &= ~16;
+
+		if (dir1 > 7) dir0_msn++;  /* M II */
+		tmp = (!(dir0_lsn & 7) || dir0_lsn & 1) ? 2 : 0;
+		Cx86_cb[tmp] = cyrix_model_mult2[dir0_lsn & 7];
+		p = Cx86_cb+tmp;
+        	if (((dir1 & 0x0f) > 4) || ((dir1 & 0xf0) == 0x20))
+			Cx86_step[0]++;
+		break;
+
+	case 0xf:  /* Cyrix 486 without DIR registers */
+		switch (dir0_lsn) {
+		case 0xd:  /* either a 486SLC or DLC w/o DEVID */
+			dir0_msn = 0;
+			p = Cx486_name[(c->hard_math) ? 1 : 0];
+			break;
 
-	cpuid(0x80000000, &n, &dummy, &dummy, &dummy);
-	if (n < 4)
-		return 0;
-	v = (unsigned int *) c->x86_model_id;
-	cpuid(0x80000002, &v[0], &v[1], &v[2], &v[3]);
-	cpuid(0x80000003, &v[4], &v[5], &v[6], &v[7]);
-	cpuid(0x80000004, &v[8], &v[9], &v[10], &v[11]);
-	c->x86_model_id[48] = 0;
-	return 1;
+		case 0xe:  /* a 486S A step */
+			dir0_msn = 0;
+			p = Cx486S_name[0];
+			strcpy(Cx86_step, cyrix_model_oldstep);
+			c->x86_mask = 1; /* must != 0 to print */
+			break;
+		break;
+		}
+	}
+	strcpy(buf, Cx86_model[dir0_msn & 7]);
+	if (p) strcat(buf, p);
+	return;
 }
 
 __initfunc(void get_cpu_vendor(struct cpuinfo_x86 *c))
@@ -545,7 +403,7 @@ __initfunc(void get_cpu_vendor(struct cpuinfo_x86 *c))
 		c->x86_vendor = X86_VENDOR_INTEL;
 	else if (!strcmp(v, "AuthenticAMD"))
 		c->x86_vendor = X86_VENDOR_AMD;
-	else if (!strncmp(v, "Cyrix", 5))
+	else if (!strcmp(v, "CyrixInstead"))
 		c->x86_vendor = X86_VENDOR_CYRIX;
 	else if (!strcmp(v, "UMC UMC UMC "))
 		c->x86_vendor = X86_VENDOR_UMC;
@@ -577,15 +435,6 @@ static struct cpu_model_info cpu_models[] __initdata = {
 	  { "Pentium Pro A-step", "Pentium Pro", NULL, "Pentium II (Klamath)", 
 	    NULL, "Pentium II (Deschutes)", NULL, NULL, NULL, NULL, NULL, NULL,
 	    NULL, NULL, NULL, NULL }},
-	{ X86_VENDOR_CYRIX,	4,
-	  { NULL, NULL, NULL, NULL, "MediaGX", NULL, NULL, NULL, NULL, "5x86",
-	    NULL, NULL, NULL, NULL, NULL, NULL }},
-	{ X86_VENDOR_CYRIX,	5,
-	  { NULL, NULL, "6x86", NULL, "GXm", NULL, NULL, NULL, NULL,
-	    NULL, NULL, NULL, NULL, NULL, NULL, NULL }},
-	{ X86_VENDOR_CYRIX,	6,
-	  { "6x86MX", NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, NULL, 
-	    NULL, NULL, NULL, NULL, NULL, NULL }},
 	{ X86_VENDOR_AMD,	4,
 	  { NULL, NULL, NULL, "DX/2", NULL, NULL, NULL, "DX/2-WB", "DX/4",
 	    "DX/4-WB", NULL, NULL, NULL, NULL, "Am5x86-WT", "Am5x86-WB" }},
@@ -618,8 +467,10 @@ __initfunc(void identify_cpu(struct cpuinfo_x86 *c))
 	    c->cpuid_level < 0)
 		return;
 
-	if (c->x86_vendor == X86_VENDOR_CYRIX && cyrix_model(c))
+	if (c->x86_vendor == X86_VENDOR_CYRIX) {
+		cyrix_model(c);
 		return;
+	}
 
 	if (c->x86_model < 16)
 		for (i=0; i<sizeof(cpu_models)/sizeof(struct cpu_model_info); i++)
@@ -659,9 +510,12 @@ __initfunc(void print_cpu_info(struct cpuinfo_x86 *c))
 	else
 		printk("%s", c->x86_model_id);
 
-	if (c->x86_mask)
-		printk(" stepping %02x", c->x86_mask);
-
+	if (c->x86_mask) {
+		if (c->x86_vendor == X86_VENDOR_CYRIX)
+			printk(" stepping %s", Cx86_step);
+		else
+			printk(" stepping %02x", c->x86_mask);
+	}
 	printk("\n");
 }
 
@@ -697,9 +551,7 @@ int get_cpuinfo(char * buffer)
 			       c->x86_vendor_id[0] ? c->x86_vendor_id : "unknown");
 		if (c->x86_mask) {
 			if (c->x86_vendor == X86_VENDOR_CYRIX)
-				p += sprintf(p, "stepping\t: %d rev %d\n",
-					     c->x86_mask >> 4,
-					     c->x86_mask & 0x0f);
+				p += sprintf(p, "stepping\t: %s\n", Cx86_step);
 			else
 				p += sprintf(p, "stepping\t: %d\n", c->x86_mask);
 		} else

drivers/char/vc_screen.c

@@ -20,8 +20,6 @@
  *	 - making it shorter - scr_readw are macros which expand in PRETTY long code
  */
 
-#include <linux/config.h>
-
 #include <linux/kernel.h>
 #include <linux/major.h>
 #include <linux/errno.h>

drivers/misc/parport_pc.c

@@ -783,11 +783,6 @@ int parport_pc_init(int *io, int *irq, int *dma)
 		count += probe_one_port(0x278, irq[0], dma[0]);
 	}
 
-	/* Give any attached devices a chance to gather their thoughts */
-	current->state = TASK_INTERRUPTIBLE;
-	current->timeout = jiffies + 75;
-	schedule ();
-
 	return count;
 }
 

drivers/misc/parport_share.c

@@ -360,9 +360,10 @@ int parport_claim(struct pardevice *dev)
 			/* First add ourselves to the end of the wait list. */
 			dev->waitnext = NULL;
 			dev->waitprev = port->waittail;
-			if (port->waittail)
+			if (port->waittail) {
 				port->waittail->waitnext = dev;
-			else
+				port->waittail = dev;
+			} else
 				port->waithead = port->waittail = dev;
 		}
 		spin_unlock_irqrestore (&port->lock, flags);

drivers/net/eepro100.c

@@ -33,7 +33,23 @@ static int rxdmacount = 0;
 
 /* Set the copy breakpoint for the copy-only-tiny-buffer Rx method.
    Lower values use more memory, but are faster. */
-static int rx_copybreak = 200;
+/*
+ * NOTE! The value of 2000 means that this optimization never gets
+ * used. Rationale: it seems to be broken when in low-memory situations,
+ * apparently when alloc_skb() can return NULL the clever list of
+ * copy-buffers can get buggered. 
+ *
+ * My personal suspicion is that the allocation failure will cause
+ * us to not remove the skb from the list of available buffers, but
+ * we'd already have done a "skb_push()" with the data we got, so
+ * the buffer stays on the list but the available memory in it
+ * shrinks until we panic.
+ *
+ * Donald, when you fix this you can shrink this value again.
+ *
+ *		Linus
+ */
+static int rx_copybreak = 2000;
 
 /* Maximum events (Rx packets, etc.) to handle at each interrupt. */
 static int max_interrupt_work = 200;

drivers/scsi/ncr53c8xx.h

@@ -42,6 +42,8 @@
 #ifndef NCR53C8XX_H
 #define NCR53C8XX_H
 
+#include <linux/config.h>
+
 #if 0
 #define CONFIG_SCSI_NCR53C8XX_PROFILE
 #endif
@@ -62,7 +64,6 @@
 #if !defined(LINUX_VERSION_CODE)
 #include <linux/version.h>
 #endif
-#include <linux/config.h>
 
 #define LinuxVersionCode(v, p, s) (((v)<<16)+((p)<<8)+(s))
 

drivers/sound/Readme.modules

@@ -29,7 +29,7 @@ modules in the usual way.
  Then, add to your /etc/modules.conf or /etc/conf.modules something like:
 
 alias char-major-14 sb
-post-install sb modprobe  "-k" "adlib_card"
+post-install sb /sbin/modprobe "-k" "adlib_card"
 options sb io=0x220 irq=7 dma=1 dma16=5 mpu_io=0x330
 options adlib_card io=0x388     # FM synthesizer
 

drivers/sound/msnd_classic.c

@@ -24,6 +24,7 @@
  *
  ********************************************************************/
 
+#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/malloc.h>
 #include <linux/types.h>

drivers/sound/msnd_pinnacle.c

@@ -24,6 +24,7 @@
  *
  ********************************************************************/
 
+#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/malloc.h>
 #include <linux/types.h>

drivers/sound/sound_core.c

@@ -34,6 +34,7 @@
  *	locking at some point in 2.3.x.
  */
 
+#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/malloc.h>
 #include <linux/types.h>

drivers/video/promfb.c

@@ -3,7 +3,6 @@
  * Copyright (C) 1998  Eddie C. Dost  (ecd@skynet.be)
  */
 
-#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/errno.h>

fs/coda/stats.c

@@ -7,7 +7,6 @@
  *
  */
 
-#include <linux/config.h>
 #include <linux/sched.h>
 #include <linux/mm.h>
 #include <linux/sysctl.h>

fs/namei.c

@@ -86,50 +86,16 @@
  * semantics.  See the comments in "open_namei" and "do_link" below.
  */
 
-char * getname_quicklist = NULL;
-int getname_quickcount = 0;
-spinlock_t getname_quicklock = SPIN_LOCK_UNLOCKED;
-
-/* Tuning: increase locality by reusing same pages again...
- * if getname_quicklist becomes too long on low memory machines, either a limit
- * should be added or after a number of cycles some pages should
- * be released again ...
- */
 static inline char * get_page(void)
 {
 	char * res;
-	spin_lock(&getname_quicklock);
-	res = getname_quicklist;
-	if (res) {
-#ifdef DEBUG
-		char * tmp = res;
-		int i;
-		for(i=0; i<getname_quickcount; i++)
-			tmp = *(char**)tmp;
-		if (tmp)
-			printk("bad quicklist %x\n", (int)tmp);
-#endif
-		getname_quicklist = *(char**)res;
-		getname_quickcount--;
-	}
-	spin_unlock(&getname_quicklock);
-	if (!res)
-		res = (char*)__get_free_page(GFP_KERNEL);
+	res = (char*)__get_free_page(GFP_KERNEL);
 	return res;
 }
 
 inline void putname(char * name)
 {
-	if (name) {
-		spin_lock(&getname_quicklock);
-		*(char**)name = getname_quicklist;
-		getname_quicklist = name;
-		getname_quickcount++;
-		spin_unlock(&getname_quicklock);
-	}
-	/* if a getname_quicklist limit is necessary to introduce, call
-	 * free_page((unsigned long) name);
-	 */
+	free_page((unsigned long) name); 
 }
 
 /* In order to reduce some races, while at the same time doing additional

fs/vfat/namei.c

@@ -1105,6 +1105,9 @@ static int vfat_find(struct inode *dir,struct qstr* qname,
 	fil.f_pos = 0;
 	vf.name = qname->name;
 	vf.len = qname->len;
+	while (vf.len && vf.name[vf.len-1] == '.') {
+		vf.len--;
+	}
 	vf.new_filename = new_filename;
 	vf.found = 0;
 	vf.posix = MSDOS_SB(sb)->options.posixfs;
@@ -1135,7 +1138,7 @@ static int vfat_find(struct inode *dir,struct qstr* qname,
 		goto cleanup;
 	}
 	
-	res = vfat_build_slots(dir, qname->name, qname->len, ds, 
+	res = vfat_build_slots(dir, qname->name, vf.len, ds, 
 			       &slots, &is_long);
 	if (res < 0) goto cleanup;
 

include/asm-i386/bugs.h

@@ -2,6 +2,11 @@
  *  include/asm-i386/bugs.h
  *
  *  Copyright (C) 1994  Linus Torvalds
+ *
+ *  Cyrix stuff, June 1998 by:
+ *	- Rafael R. Reilova (moved everything from head.S),
+ *	- Channing Corn (tests & fixes),
+ *	- Andrew D. Balsa (code cleanup).
  */
 
 /*
@@ -226,13 +231,99 @@ __initfunc(static void check_pentium_f00f(void))
 	}
 }
 
+/*
+ * perform the Cyrix 5/2 test (!0 means it's a Cyrix)
+ */
+
+static inline int test_cyrix_52div(void)
+{
+	int test;
+
+	__asm__ __volatile__("xor %%eax,%%eax\n\t"
+	     "sahf\n\t"
+	     "movb $5,%%al\n\t"
+	     "movb $2,%%bl\n\t"
+	     "div %%bl\n\t"
+	     "lahf\n\t"
+	     "andl $0x200,%%eax": "=a" (test) : : "bx", "cc");
+
+	return test;
+}
+
+/*
+ * Cyrix CPUs without cpuid or with cpuid not yet enabled can be detected
+ * by the fact that they preserve the flags across the division of 5/2.
+ * PII and PPro exhibit this behavior too, but they have cpuid available.
+ */
+
+__initfunc(static void check_cyrix_cpu(void))
+{
+	if (boot_cpu_data.cpuid_level == -1 && boot_cpu_data.x86 == 4
+	    && test_cyrix_52div()) {
+
+		/* default to an unknown Cx486, (we will diferentiate later) */
+		/* NOTE:  using 0xff since 0x00 is a valid DIR0 value */
+		strcpy(boot_cpu_data.x86_vendor_id, "CyrixInstead");
+		boot_cpu_data.x86_model = 0xff;
+		boot_cpu_data.x86_mask = 0;
+	}
+}
+
+/*
+ * Fix two problems with the Cyrix 686 and 686L:
+ *   -- the cpuid is disabled on power up, enable it, use it.
+ *   -- the SLOP bit needs resetting on some motherboards due to old BIOS,
+ *      so that the udelay loop calibration works well.  Recalibrate.
+ */
+
+extern void calibrate_delay(void) __init;
+
+__initfunc(static void check_cx686_cpuid_slop(void))
+{
+	if (boot_cpu_data.x86_vendor == X86_VENDOR_CYRIX &&
+	    (boot_cpu_data.x86_model & 0xf0) == 0x30) {  /* 686(L) */
+		int dummy;
+		unsigned char ccr3, ccr5;
+
+		cli();
+		ccr3 = getCx86(CX86_CCR3);
+		setCx86(CX86_CCR3, (ccr3 & 0x0f) | 0x10);      /* enable MAPEN  */
+		setCx86(CX86_CCR4, getCx86(CX86_CCR4) | 0x80); /* enable cpuid  */
+		ccr5 = getCx86(CX86_CCR5);
+		if (ccr5 & 2)   /* reset SLOP if needed, old BIOS do this wrong */
+			setCx86(CX86_CCR5, ccr5 & 0xfd);
+		setCx86(CX86_CCR3, ccr3);                      /* disable MAPEN */
+		sti();
+
+		boot_cpu_data.cpuid_level = 1;  /* should cover all 6x86(L) */
+		boot_cpu_data.x86 = 5;
+
+		/* we know we have level 1 available on the 6x86(L) */
+		cpuid(1, &dummy, &dummy, &dummy,
+		      &boot_cpu_data.x86_capability);
+		/*
+		 * DON'T use the x86_mask and x86_model from cpuid, these are
+		 * not as accurate (or the same) as those from the DIR regs.
+		 * already in place after cyrix_model() in setup.c
+		 */
+
+		if (ccr5 & 2) { /* possible wrong calibration done */
+			printk(KERN_INFO "Recalibrating delay loop with SLOP bit reset\n");
+			calibrate_delay();
+			boot_cpu_data.loops_per_sec = loops_per_sec;
+		}
+	}
+}
+
 __initfunc(static void check_bugs(void))
 {
+	check_cyrix_cpu();
 	identify_cpu(&boot_cpu_data);
 #ifndef __SMP__
 	printk("CPU: ");
 	print_cpu_info(&boot_cpu_data);
 #endif
+	check_cx686_cpuid_slop();
 	check_tlb();
 	check_fpu();
 	check_hlt();

include/asm-i386/processor.h

@@ -99,6 +99,51 @@ extern char ignore_irq13;
 extern void identify_cpu(struct cpuinfo_x86 *);
 extern void print_cpu_info(struct cpuinfo_x86 *);
 
+/*
+ *	Generic CPUID function
+ */
+extern inline void cpuid(int op, int *eax, int *ebx, int *ecx, int *edx)
+{
+	__asm__("cpuid"
+		: "=a" (*eax),
+		  "=b" (*ebx),
+		  "=c" (*ecx),
+		  "=d" (*edx)
+		: "a" (op)
+		: "cc");
+}
+
+/*
+ *      Cyrix CPU configuration register indexes
+ */
+#define CX86_CCR2 0xc2
+#define CX86_CCR3 0xc3
+#define CX86_CCR4 0xe8
+#define CX86_CCR5 0xe9
+#define CX86_DIR0 0xfe
+#define CX86_DIR1 0xff
+
+/*
+ *      Cyrix CPU indexed register access macros
+ */
+
+extern inline unsigned char getCx86(unsigned char reg)
+{
+	unsigned char data;
+
+	__asm__ __volatile__("movb %1,%%al\n\t"
+		      "outb %%al,$0x22\n\t"
+		      "inb $0x23,%%al" : "=a" (data) : "q" (reg));
+	return data;
+}
+
+extern inline void setCx86(unsigned char reg, unsigned char data)
+{
+	__asm__ __volatile__("outb %%al,$0x22\n\t"
+	     "movb %1,%%al\n\t"
+	     "outb %%al,$0x23" : : "a" (reg), "q" (data));
+}
+
 /*
  * Bus types (default is ISA, but people can check others with these..)
  */
@@ -248,13 +293,8 @@ extern inline unsigned long thread_saved_pc(struct thread_struct *t)
 	return ((unsigned long *)t->esp)[3];
 }
 
-/* Allocation and freeing of basic task resources. */
-/*
- * NOTE! The task struct and the stack go together
- */
-#define alloc_task_struct() \
-	((struct task_struct *) __get_free_pages(GFP_KERNEL,1))
-#define free_task_struct(p)	free_pages((unsigned long)(p),1)
+extern struct task_struct * alloc_task_struct(void);
+extern void free_task_struct(struct task_struct *);
 
 #define init_task	(init_task_union.task)
 #define init_stack	(init_task_union.stack)

include/linux/console_struct.h

@@ -10,8 +10,6 @@
  */
 #define CUR_DEFAULT CUR_UNDERLINE
 
-#include <linux/config.h>
-
 #define NPAR 16
 
 struct vc_data {

include/linux/mm.h

@@ -254,21 +254,9 @@ extern inline unsigned long get_free_page(int gfp_mask)
 /* memory.c & swap.c*/
 
 /*
- * This traverses "nr" memory size lists,
- * and returns true if there is enough memory.
- *
- * For example, we want to keep on waking up
- * kswapd every once in a while until the highest
- * memory order has an entry (ie nr == 0), but
- * we want to do it in the background.
- *
- * We want to do it in the foreground only if
- * none of the three highest lists have enough
- * memory. Random number.
+ * Decide if we should try to do some swapout..
  */
-extern int free_memory_available(int nr);
-#define kswapd_continue()	(!free_memory_available(3))
-#define kswapd_wakeup()		(!free_memory_available(0))
+extern int free_memory_available(void);
 
 #define free_page(addr) free_pages((addr),0)
 extern void FASTCALL(free_pages(unsigned long addr, unsigned long order));
@@ -330,8 +318,8 @@ extern void put_cached_page(unsigned long);
 #define GFP_BUFFER	(__GFP_LOW | __GFP_WAIT)
 #define GFP_ATOMIC	(__GFP_HIGH)
 #define GFP_USER	(__GFP_LOW | __GFP_WAIT | __GFP_IO)
-#define GFP_KERNEL	(__GFP_LOW | __GFP_WAIT | __GFP_IO)
-#define GFP_NFS		(__GFP_MED | __GFP_WAIT | __GFP_IO)
+#define GFP_KERNEL	(__GFP_MED | __GFP_WAIT | __GFP_IO)
+#define GFP_NFS		(__GFP_HIGH | __GFP_WAIT | __GFP_IO)
 
 /* Flag - indicates that the buffer will be suitable for DMA.  Ignored on some
    platforms, used as appropriate on others */

include/linux/swap.h

@@ -50,7 +50,7 @@ struct sysinfo;
 extern int shm_swap (int, int);
 
 /* linux/mm/vmscan.c */
-extern int try_to_free_page(int);
+extern int try_to_free_pages(unsigned int gfp_mask, int count);
 
 /* linux/mm/page_io.c */
 extern void rw_swap_page(int, unsigned long, char *, int);
@@ -128,21 +128,6 @@ static inline int is_page_shared(struct page *page)
 	return (count > 1);
 }
 
-/*
- * When we're freeing pages from a user application, we want
- * to cluster swapouts too.	-- Rik.
- * linux/mm/page_alloc.c
- */
-static inline int try_to_free_pages(int gfp_mask, int count)
-{
-	int retval = 0;
-	while (count--) {
-		if (try_to_free_page(gfp_mask))
-			retval = 1;
-	}
-	return retval;
-}
-
 /*
  * Make these inline later once they are working properly.
  */

include/linux/vt_buffer.h

@@ -16,7 +16,7 @@
 #include <linux/config.h>
 
 #ifdef CONFIG_VGA_CONSOLE
-#ifndef CONFIG_FB_MODULE
+#if !defined(CONFIG_FB) && !defined(CONFIG_FB_MODULE)
 #define VT_BUF_VRAM_ONLY
 #endif
 #include <asm/vga.h>

include/linux/vt_kern.h

@@ -6,6 +6,7 @@
  * with information needed by the vt package
  */
 
+#include <linux/config.h>
 #include <linux/vt.h>
 
 /*

kernel/kmod.c

@@ -21,7 +21,7 @@
 	modprobe_path is set via /proc/sys.
 */
 char modprobe_path[256] = "/sbin/modprobe";
-static char * envp[] = { "HOME=/", "TERM=linux", "PATH=/usr/bin:/bin", NULL };
+static char * envp[] = { "HOME=/", "TERM=linux", "PATH=/sbin:/usr/sbin:/bin:/usr/bin", NULL };
 
 /*
 	exec_modprobe is spawned from a kernel-mode user process,

mm/filemap.c

@@ -125,8 +125,8 @@ int shrink_mmap(int priority, int gfp_mask)
 	struct buffer_head *tmp, *bh;
 	int count_max, count_min;
 
-	count_max = (limit<<1) >> (priority>>1);
-	count_min = (limit<<1) >> (priority);
+	count_max = (limit<<2) >> (priority>>1);
+	count_min = (limit<<2) >> (priority);
 
 	page = mem_map + clock;
 	do {

mm/mmap.c

@@ -57,19 +57,19 @@ int vm_enough_memory(long pages)
 	 * simple, it hopefully works in most obvious cases.. Easy to
 	 * fool it, but this should catch most mistakes.
 	 */
-	long freepages;
+	long free;
 	
         /* Sometimes we want to use more memory than we have. */
 	if (sysctl_overcommit_memory)
 	    return 1;
 
-	freepages = buffermem >> PAGE_SHIFT;
-	freepages += page_cache_size;
-	freepages >>= 1;
-	freepages += nr_free_pages;
-	freepages += nr_swap_pages;
-	freepages -= num_physpages >> 4;
-	return freepages > pages;
+	free = buffermem >> PAGE_SHIFT;
+	free += page_cache_size;
+	free >>= 1;
+	free += nr_free_pages;
+	free += nr_swap_pages;
+	free -= num_physpages >> 4;
+	return free > pages;
 }
 
 /* Remove one vm structure from the inode's i_mmap ring. */

mm/page_alloc.c

@@ -103,48 +103,28 @@ spinlock_t page_alloc_lock = SPIN_LOCK_UNLOCKED;
 /*
  * This routine is used by the kernel swap daemon to determine
  * whether we have "enough" free pages. It is fairly arbitrary,
- * but this had better return false if any reasonable "get_free_page()"
- * allocation could currently fail..
+ * having a low-water and high-water mark.
  *
- * This will return zero if no list was found, non-zero
- * if there was memory (the bigger, the better).
+ * This returns:
+ *  0 - urgent need for memory
+ *  1 - need some memory, but do it slowly in the background
+ *  2 - no need to even think about it.
  */
-int free_memory_available(int nr)
+int free_memory_available(void)
 {
-	int retval = 0;
-	unsigned long flags;
-	struct free_area_struct * list;
+	static int available = 1;
 
-	/*
-	 * If we have more than about 3% to 5% of all memory free,
-	 * consider it to be good enough for anything.
-	 * It may not be, due to fragmentation, but we
-	 * don't want to keep on forever trying to find
-	 * free unfragmented memory.
-	 * Added low/high water marks to avoid thrashing -- Rik.
-	 */
-	if (nr_free_pages > (nr ? freepages.low : freepages.high))
-		return nr+1;
+	if (nr_free_pages < freepages.low) {
+		available = 0;
+		return 0;
+	}
 
-	list = free_area + NR_MEM_LISTS;
-	spin_lock_irqsave(&page_alloc_lock, flags);
-	/* We fall through the loop if the list contains one
-	 * item. -- thanks to Colin Plumb <colin@nyx.net>
-	 */
-	do {
-		list--;
-		/* Empty list? Bad - we need more memory */
-		if (list->next == memory_head(list))
-			break;
-		/* One item on the list? Look further */
-		if (list->next->next == memory_head(list))
-			continue;
-		/* More than one item? We're ok */
-		retval = nr + 1;
-		break;
-	} while (--nr >= 0);
-	spin_unlock_irqrestore(&page_alloc_lock, flags);
-	return retval;
+	if (nr_free_pages > freepages.high) {
+		available = 1;
+		return 2;
+	}
+
+	return available;
 }
 
 static inline void free_pages_ok(unsigned long map_nr, unsigned long order)
@@ -217,13 +197,11 @@ void free_pages(unsigned long addr, unsigned long order)
 	change_bit((index) >> (1+(order)), (area)->map)
 #define CAN_DMA(x) (PageDMA(x))
 #define ADDRESS(x) (PAGE_OFFSET + ((x) << PAGE_SHIFT))
-#define RMQUEUE(order, maxorder, dma) \
+#define RMQUEUE(order, dma) \
 do { struct free_area_struct * area = free_area+order; \
      unsigned long new_order = order; \
 	do { struct page *prev = memory_head(area), *ret = prev->next; \
 		while (memory_head(area) != ret) { \
-			if (new_order >= maxorder && ret->next == prev) \
-				break; \
 			if (!dma || CAN_DMA(ret)) { \
 				unsigned long map_nr = ret->map_nr; \
 				(prev->next = ret->next)->prev = prev; \
@@ -255,39 +233,36 @@ do { unsigned long size = 1 << high; \
 
 unsigned long __get_free_pages(int gfp_mask, unsigned long order)
 {
-	unsigned long flags, maxorder;
+	unsigned long flags;
 
 	if (order >= NR_MEM_LISTS)
 		goto nopage;
 
-	/*
-	 * "maxorder" is the highest order number that we're allowed
-	 * to empty in order to find a free page..
-	 */
-	maxorder = NR_MEM_LISTS-1;
-	if (gfp_mask & __GFP_HIGH)
-		maxorder = NR_MEM_LISTS;
-
-	if (in_interrupt() && (gfp_mask & __GFP_WAIT)) {
-		static int count = 0;
-		if (++count < 5) {
-			printk("gfp called nonatomically from interrupt %p\n",
-				__builtin_return_address(0));
-			gfp_mask &= ~__GFP_WAIT;
+	if (gfp_mask & __GFP_WAIT) {
+		if (in_interrupt()) {
+			static int count = 0;
+			if (++count < 5) {
+				printk("gfp called nonatomically from interrupt %p\n",
+					__builtin_return_address(0));
+			}
+			goto nopage;
 		}
-	}
 
-	for (;;) {
-		spin_lock_irqsave(&page_alloc_lock, flags);
-		RMQUEUE(order, maxorder, (gfp_mask & GFP_DMA));
-		spin_unlock_irqrestore(&page_alloc_lock, flags);
-		if (!(gfp_mask & __GFP_WAIT))
-			break;
-		if (!try_to_free_pages(gfp_mask, SWAP_CLUSTER_MAX))
-			break;
-		gfp_mask &= ~__GFP_WAIT;	/* go through this only once */
-		maxorder = NR_MEM_LISTS;	/* Allow anything this time */
+		if (freepages.min > nr_free_pages) {
+			int freed;
+			freed = try_to_free_pages(gfp_mask, SWAP_CLUSTER_MAX);
+			/*
+			 * Low priority (user) allocations must not
+			 * succeed if we didn't have enough memory
+			 * and we couldn't get more..
+			 */
+			if (!freed && !(gfp_mask & (__GFP_MED | __GFP_HIGH)))
+				goto nopage;
+		}
 	}
+	spin_lock_irqsave(&page_alloc_lock, flags);
+	RMQUEUE(order, (gfp_mask & GFP_DMA));
+	spin_unlock_irqrestore(&page_alloc_lock, flags);
 nopage:
 	return 0;
 }
@@ -303,6 +278,11 @@ void show_free_areas(void)
  	unsigned long total = 0;
 
 	printk("Free pages:      %6dkB\n ( ",nr_free_pages<<(PAGE_SHIFT-10));
+	printk("Free: %d (%d %d %d)\n",
+		nr_free_pages,
+		freepages.min,
+		freepages.low,
+		freepages.high);
 	spin_lock_irqsave(&page_alloc_lock, flags);
  	for (order=0 ; order < NR_MEM_LISTS; order++) {
 		struct page * tmp;

mm/vmscan.c

@@ -444,7 +444,7 @@ static int swap_out(unsigned int priority, int gfp_mask)
  * to be.  This works out OK, because we now do proper aging on page
  * contents. 
  */
-static inline int do_try_to_free_page(int gfp_mask)
+static int do_try_to_free_page(int gfp_mask)
 {
 	static int state = 0;
 	int i=6;
@@ -457,9 +457,10 @@ static inline int do_try_to_free_page(int gfp_mask)
 	stop = 3;
 	if (gfp_mask & __GFP_WAIT)
 		stop = 0;
+
 	if (((buffermem >> PAGE_SHIFT) * 100 > buffer_mem.borrow_percent * num_physpages)
 		   || (page_cache_size * 100 > page_cache.borrow_percent * num_physpages))
-		state = 0;
+		shrink_mmap(i, gfp_mask);
 
 	switch (state) {
 		do {
@@ -484,23 +485,6 @@ static inline int do_try_to_free_page(int gfp_mask)
 	return 0;
 }
 
-/*
- * This is REALLY ugly.
- *
- * We need to make the locks finer granularity, but right
- * now we need this so that we can do page allocations
- * without holding the kernel lock etc.
- */
-int try_to_free_page(int gfp_mask)
-{
-	int retval;
-
-	lock_kernel();
-	retval = do_try_to_free_page(gfp_mask);
-	unlock_kernel();
-	return retval;
-}
-
 /*
  * Before we start the kernel thread, print out the 
  * kswapd initialization message (otherwise the init message 
@@ -573,15 +557,16 @@ int kswapd(void *unused)
 		 * woken up more often and the rate will be even
 		 * higher).
 		 */
-		tries = pager_daemon.tries_base >> free_memory_available(3);
+		tries = pager_daemon.tries_base;
+		tries >>= 4*free_memory_available();
 	
 		while (tries--) {
 			int gfp_mask;
 
-			if (++tried > pager_daemon.tries_min && free_memory_available(0))
+			if (free_memory_available() > 1)
 				break;
 			gfp_mask = __GFP_IO;
-			try_to_free_page(gfp_mask);
+			do_try_to_free_page(gfp_mask);
 			/*
 			 * Syncing large chunks is faster than swapping
 			 * synchronously (less head movement). -- Rik.
@@ -597,6 +582,27 @@ int kswapd(void *unused)
 	return 0;
 }
 
+/*
+ * This is REALLY ugly.
+ *
+ * We need to make the locks finer granularity, but right
+ * now we need this so that we can do page allocations
+ * without holding the kernel lock etc.
+ */
+int try_to_free_pages(unsigned int gfp_mask, int count)
+{
+	int retval;
+
+	lock_kernel();
+	retval = 0;
+	do {
+		retval |= do_try_to_free_page(gfp_mask);
+		count--;
+	} while (count > 0);
+	unlock_kernel();
+	return retval;
+}
+
 /* 
  * The swap_tick function gets called on every clock tick.
  */
@@ -616,11 +622,11 @@ void swap_tick(void)
 	 * Schedule for wakeup if there isn't lots
 	 * of free memory.
 	 */
-	switch (free_memory_available(3)) {
+	switch (free_memory_available()) {
 	case 0:
 		want = now;
 		/* Fall through */
-	case 1 ... 3:
+	case 1:
 		want_wakeup = 1;
 	default:
 	}