5 * - General arch-specific stuff
8 #include <threads_int.h>
13 #define DEBUG_TO_SERIAL 1
14 #define SERIAL_PORT 0x3F8
15 #define GDB_SERIAL_PORT 0x2F8
19 extern struct sShortSpinlock glDebug_Lock;
20 extern tMutex glPhysAlloc;
21 #define TRACE_LOCK_COND (Lock != &glDebug_Lock && Lock != &glThreadListLock && Lock != &glPhysAlloc.Protector)
22 //#define TRACE_LOCK_COND (Lock != &glDebug_Lock && Lock != &glPhysAlloc.Protector)
25 extern int GetCPUNum(void);
28 Uint64 __divmod64(Uint64 Num, Uint64 Den, Uint64 *Rem);
29 Uint64 __udivdi3(Uint64 Num, Uint64 Den);
30 Uint64 __umoddi3(Uint64 Num, Uint64 Den);
33 int gbDebug_SerialSetup = 0;
34 int gbGDB_SerialSetup = 0;
38 * \brief Determine if a short spinlock is locked
39 * \param Lock Lock pointer
41 int IS_LOCKED(struct sShortSpinlock *Lock)
47 * \brief Check if the current CPU has the lock
48 * \param Lock Lock pointer
50 int CPU_HAS_LOCK(struct sShortSpinlock *Lock)
52 return Lock->Lock == GetCPUNum() + 1;
56 * \brief Acquire a Short Spinlock
57 * \param Lock Lock pointer
59 * This type of mutex should only be used for very short sections of code,
60 * or in places where a Mutex_* would be overkill, such as appending
61 * an element to linked list (usually two assignement lines in C)
63 * \note This type of lock halts interrupts, so ensure that no timing
64 * functions are called while it is held. As a matter of fact, spend as
65 * little time as possible with this lock held
66 * \note If \a STACKED_LOCKS is set, this type of spinlock can be nested
68 void SHORTLOCK(struct sShortSpinlock *Lock)
71 int cpu = GetCPUNum() + 1;
73 // Save interrupt state
74 __ASM__ ("pushf;\n\tpop %0" : "=r"(IF));
75 IF &= 0x200; // AND out all but the interrupt flag
80 //Log_Log("LOCK", "%p locked by %p", Lock, __builtin_return_address(0));
81 Debug("%i %p obtaining %p (Called by %p)", cpu-1, __builtin_return_address(0), Lock, __builtin_return_address(1));
87 // Wait for another CPU to release
90 "xor %%eax, %%eax;\n\t"
91 "lock cmpxchgl %0, (%1);\n\t"
93 :: "r"(cpu), "r"(&Lock->Lock)
94 : "eax" // EAX clobbered
100 if( TRACE_LOCK_COND )
102 //Log_Log("LOCK", "%p locked by %p", Lock, __builtin_return_address(0));
103 Debug("%i %p locked by %p\t%p", cpu-1, Lock, __builtin_return_address(0), __builtin_return_address(1));
109 * \brief Release a short lock
110 * \param Lock Lock pointer
112 void SHORTREL(struct sShortSpinlock *Lock)
115 if( TRACE_LOCK_COND )
117 //Log_Log("LOCK", "%p released by %p", Lock, __builtin_return_address(0));
118 Debug("Lock %p released by %p\t%p", Lock, __builtin_return_address(0), __builtin_return_address(1));
122 // Lock->IF can change anytime once Lock->Lock is zeroed
134 int putDebugChar(char ch)
136 if(!gbGDB_SerialSetup) {
137 outb(GDB_SERIAL_PORT + 1, 0x00); // Disable all interrupts
138 outb(GDB_SERIAL_PORT + 3, 0x80); // Enable DLAB (set baud rate divisor)
139 outb(GDB_SERIAL_PORT + 0, 0x0C); // Set divisor to 12 (lo byte) 9600 baud
140 outb(GDB_SERIAL_PORT + 1, 0x00); // (base is (hi byte)
141 outb(GDB_SERIAL_PORT + 3, 0x03); // 8 bits, no parity, one stop bit (8N1)
142 outb(GDB_SERIAL_PORT + 2, 0xC7); // Enable FIFO with 14-byte threshold and clear it
143 outb(GDB_SERIAL_PORT + 4, 0x0B); // IRQs enabled, RTS/DSR set
144 gbGDB_SerialSetup = 1;
146 while( (inb(GDB_SERIAL_PORT + 5) & 0x20) == 0 );
147 outb(GDB_SERIAL_PORT, ch);
150 int getDebugChar(void)
152 if(!gbGDB_SerialSetup) {
153 outb(GDB_SERIAL_PORT + 1, 0x00); // Disable all interrupts
154 outb(GDB_SERIAL_PORT + 3, 0x80); // Enable DLAB (set baud rate divisor)
155 outb(GDB_SERIAL_PORT + 0, 0x0C); // Set divisor to 12 (lo byte) 9600 baud
156 outb(GDB_SERIAL_PORT + 1, 0x00); // (hi byte)
157 outb(GDB_SERIAL_PORT + 3, 0x03); // 8 bits, no parity, one stop bit
158 outb(GDB_SERIAL_PORT + 2, 0xC7); // Enable FIFO with 14-byte threshold and clear it
159 outb(GDB_SERIAL_PORT + 4, 0x0B); // IRQs enabled, RTS/DSR set
160 gbGDB_SerialSetup = 1;
162 while( (inb(GDB_SERIAL_PORT + 5) & 1) == 0) ;
163 return inb(GDB_SERIAL_PORT);
165 #endif /* USE_GDB_STUB */
167 void Debug_PutCharDebug(char ch)
170 __asm__ __volatile__ ( "outb %%al, $0xe9" :: "a"(((Uint8)ch)) );
174 if(!gbDebug_SerialSetup) {
175 outb(SERIAL_PORT + 1, 0x00); // Disable all interrupts
176 outb(SERIAL_PORT + 3, 0x80); // Enable DLAB (set baud rate divisor)
177 outb(SERIAL_PORT + 0, 0x0C); // Set divisor to 12 (lo byte) 9600 baud
178 outb(SERIAL_PORT + 1, 0x00); // (hi byte)
179 outb(SERIAL_PORT + 3, 0x03); // 8 bits, no parity, one stop bit
180 outb(SERIAL_PORT + 2, 0xC7); // Enable FIFO with 14-byte threshold and clear it
181 outb(SERIAL_PORT + 4, 0x0B); // IRQs enabled, RTS/DSR set
182 gbDebug_SerialSetup = 1;
184 while( (inb(SERIAL_PORT + 5) & 0x20) == 0 );
185 outb(SERIAL_PORT, ch);
189 void Debug_PutStringDebug(const char *String)
192 Debug_PutCharDebug(*String++);
195 // === IO Commands ===
196 void outb(Uint16 Port, Uint8 Data)
198 __asm__ __volatile__ ("outb %%al, %%dx"::"d"(Port),"a"(Data));
200 void outw(Uint16 Port, Uint16 Data)
202 __asm__ __volatile__ ("outw %%ax, %%dx"::"d"(Port),"a"(Data));
204 void outd(Uint16 Port, Uint32 Data)
206 __asm__ __volatile__ ("outl %%eax, %%dx"::"d"(Port),"a"(Data));
208 Uint8 inb(Uint16 Port)
211 __asm__ __volatile__ ("inb %%dx, %%al":"=a"(ret):"d"(Port));
214 Uint16 inw(Uint16 Port)
217 __asm__ __volatile__ ("inw %%dx, %%ax":"=a"(ret):"d"(Port));
220 Uint32 ind(Uint16 Port)
223 __asm__ __volatile__ ("inl %%dx, %%eax":"=a"(ret):"d"(Port));
228 * \fn void *memset(void *Dest, int Val, size_t Num)
229 * \brief Do a byte granuality set of Dest
231 void *memset(void *Dest, int Val, size_t Num)
233 Uint32 val = Val&0xFF;
236 __asm__ __volatile__ (
240 :: "D" (Dest), "a" (val), "c" (Num/4), "r" (Num&3));
244 * \brief Set double words
246 void *memsetd(void *Dest, Uint32 Val, size_t Num)
248 __asm__ __volatile__ ("rep stosl" :: "D" (Dest), "a" (Val), "c" (Num));
253 * \fn int memcmp(const void *m1, const void *m2, size_t Num)
254 * \brief Compare two pieces of memory
256 int memcmp(const void *m1, const void *m2, size_t Num)
258 const Uint8 *d1 = m1;
259 const Uint8 *d2 = m2;
260 if( Num == 0 ) return 0; // No bytes are always identical
273 * \fn void *memcpy(void *Dest, const void *Src, size_t Num)
274 * \brief Copy \a Num bytes from \a Src to \a Dest
276 void *memcpy(void *Dest, const void *Src, size_t Num)
278 tVAddr dst = (tVAddr)Dest;
279 tVAddr src = (tVAddr)Src;
280 if( (dst & 3) != (src & 3) )
282 __asm__ __volatile__ ("rep movsb" :: "D" (dst), "S" (src), "c" (Num));
283 // Debug("\nmemcpy:Num=0x%x by %p (UA)", Num, __builtin_return_address(0));
286 else if( Num > 128 && (dst & 15) == (src & 15) )
288 char tmp[16+15]; // Note, this is a hack to save/restor xmm0
289 int count = 16 - (dst & 15);
290 // Debug("\nmemcpy:Num=0x%x by %p (SSE)", Num, __builtin_return_address(0));
294 __asm__ __volatile__ ("rep movsb" : "=D"(dst),"=S"(src): "0"(dst), "1"(src), "c"(count));
298 __asm__ __volatile__ (
299 "movdqa 0(%5), %%xmm0;\n\t"
301 "movdqa 0(%1), %%xmm0;\n\t"
302 "movdqa %%xmm0, 0(%0);\n\t"
306 "movdqa %%xmm0, 0(%5);\n\t"
307 : "=r"(dst),"=r"(src)
308 : "0"(dst), "1"(src), "c"(count), "r" (((tVAddr)tmp+15)&~15)
313 __asm__ __volatile__ ("rep movsb" :: "D"(dst), "S"(src), "c"(count));
318 // Debug("\nmemcpy:Num=0x%x by %p", Num, __builtin_return_address(0));
319 __asm__ __volatile__ (
323 :: "D" (Dest), "S" (Src), "c" (Num/4), "r" (Num&3));
329 * \fn void *memcpyd(void *Dest, const void *Src, size_t Num)
330 * \brief Copy \a Num DWORDs from \a Src to \a Dest
332 void *memcpyd(void *Dest, const void *Src, size_t Num)
334 __asm__ __volatile__ ("rep movsl" :: "D" (Dest), "S" (Src), "c" (Num));
338 #include "../helpers.h"
342 Uint64 DivMod64U(Uint64 Num, Uint64 Div, Uint64 *Rem)
344 if( Div < 0x100000000ULL && Num < 0xFFFFFFFF * Div ) {
346 __asm__ __volatile__(
348 : "=a" (ret_32), "=d" (rem)
349 : "a" ( (Uint32)(Num & 0xFFFFFFFF) ), "d" ((Uint32)(Num >> 32)), "r" (Div)
355 return __divmod64(Num, Div, Rem);
359 * \fn Uint64 __udivdi3(Uint64 Num, Uint64 Den)
360 * \brief Divide two 64-bit integers
362 Uint64 __udivdi3(Uint64 Num, Uint64 Den)
365 __asm__ __volatile__ ("int $0x0");
369 if(Num <= 0xFFFFFFFF && Den <= 0xFFFFFFFF)
370 return (Uint32)Num / (Uint32)Den;
371 if(Den == 1) return Num;
372 if(Den == 2) return Num >> 1; // Speed Hacks
373 if(Den == 4) return Num >> 2; // Speed Hacks
374 if(Den == 8) return Num >> 3; // Speed Hacks
375 if(Den == 16) return Num >> 4; // Speed Hacks
376 if(Den == 32) return Num >> 5; // Speed Hacks
377 if(Den == 1024) return Num >> 10; // Speed Hacks
378 if(Den == 2048) return Num >> 11; // Speed Hacks
379 if(Den == 4096) return Num >> 12;
380 if(Num < Den) return 0;
381 if(Num < Den*2) return 1;
382 if(Num == Den*2) return 2;
384 return __divmod64(Num, Den, NULL);
388 * \fn Uint64 __umoddi3(Uint64 Num, Uint64 Den)
389 * \brief Get the modulus of two 64-bit integers
391 Uint64 __umoddi3(Uint64 Num, Uint64 Den)
395 __asm__ __volatile__ ("int $0x0"); // Call Div by Zero Error
398 if(Den == 1) return 0; // Speed Hacks
399 if(Den == 2) return Num & 1; // Speed Hacks
400 if(Den == 4) return Num & 3; // Speed Hacks
401 if(Den == 8) return Num & 7; // Speed Hacks
402 if(Den == 16) return Num & 15; // Speed Hacks
403 if(Den == 32) return Num & 31; // Speed Hacks
404 if(Den == 1024) return Num & 1023; // Speed Hacks
405 if(Den == 2048) return Num & 2047; // Speed Hacks
406 if(Den == 4096) return Num & 4095; // Speed Hacks
408 if(Num >> 32 == 0 && Den >> 32 == 0)
409 return (Uint32)Num % (Uint32)Den;
411 __divmod64(Num, Den, &ret);
417 EXPORT(memcpy); EXPORT(memset);
419 //EXPORT(memcpyw); EXPORT(memsetw);
420 EXPORT(memcpyd); EXPORT(memsetd);
421 EXPORT(inb); EXPORT(inw); EXPORT(ind);
422 EXPORT(outb); EXPORT(outw); EXPORT(outd);
423 EXPORT(__udivdi3); EXPORT(__umoddi3);