5 * - General arch-specific stuff
8 #include <threads_int.h>
10 #include <hal_proc.h> // GetCPUNum
15 #define DEBUG_TO_SERIAL 1
16 #define SERIAL_PORT 0x3F8
17 #define GDB_SERIAL_PORT 0x2F8
21 extern struct sShortSpinlock glDebug_Lock;
22 extern tMutex glPhysAlloc;
23 #define TRACE_LOCK_COND (Lock != &glDebug_Lock && Lock != &glThreadListLock && Lock != &glPhysAlloc.Protector)
24 //#define TRACE_LOCK_COND (Lock != &glDebug_Lock && Lock != &glPhysAlloc.Protector)
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;
55 void __AtomicTestSetLoop(Uint *Ptr, Uint Value)
59 "xor %%eax, %%eax;\n\t"
60 "lock cmpxchgl %0, (%1);\n\t"
62 :: "r"(Value), "r"(Ptr)
63 : "eax" // EAX clobbered
67 * \brief Acquire a Short Spinlock
68 * \param Lock Lock pointer
70 * This type of mutex should only be used for very short sections of code,
71 * or in places where a Mutex_* would be overkill, such as appending
72 * an element to linked list (usually two assignement lines in C)
74 * \note This type of lock halts interrupts, so ensure that no timing
75 * functions are called while it is held. As a matter of fact, spend as
76 * little time as possible with this lock held
77 * \note If \a STACKED_LOCKS is set, this type of spinlock can be nested
79 void SHORTLOCK(struct sShortSpinlock *Lock)
82 int cpu = GetCPUNum() + 1;
84 // Save interrupt state
85 __ASM__ ("pushf;\n\tpop %0" : "=r"(IF));
86 IF &= 0x200; // AND out all but the interrupt flag
91 //Log_Log("LOCK", "%p locked by %p", Lock, __builtin_return_address(0));
92 Debug("%i %p obtaining %p (Called by %p)", cpu-1, __builtin_return_address(0), Lock, __builtin_return_address(1));
98 // Wait for another CPU to release
99 __AtomicTestSetLoop( (Uint*)&Lock->Lock, cpu );
103 if( TRACE_LOCK_COND )
105 //Log_Log("LOCK", "%p locked by %p", Lock, __builtin_return_address(0));
106 Debug("%i %p locked by %p\t%p", cpu-1, Lock, __builtin_return_address(0), __builtin_return_address(1));
112 * \brief Release a short lock
113 * \param Lock Lock pointer
115 void SHORTREL(struct sShortSpinlock *Lock)
118 if( TRACE_LOCK_COND )
120 //Log_Log("LOCK", "%p released by %p", Lock, __builtin_return_address(0));
121 Debug("Lock %p released by %p\t%p", Lock, __builtin_return_address(0), __builtin_return_address(1));
125 // Lock->IF can change anytime once Lock->Lock is zeroed
137 int putDebugChar(char ch)
139 if(!gbGDB_SerialSetup) {
140 outb(GDB_SERIAL_PORT + 1, 0x00); // Disable all interrupts
141 outb(GDB_SERIAL_PORT + 3, 0x80); // Enable DLAB (set baud rate divisor)
142 outb(GDB_SERIAL_PORT + 0, 0x0C); // Set divisor to 12 (lo byte) 9600 baud
143 outb(GDB_SERIAL_PORT + 1, 0x00); // (base is (hi byte)
144 outb(GDB_SERIAL_PORT + 3, 0x03); // 8 bits, no parity, one stop bit (8N1)
145 outb(GDB_SERIAL_PORT + 2, 0xC7); // Enable FIFO with 14-byte threshold and clear it
146 outb(GDB_SERIAL_PORT + 4, 0x0B); // IRQs enabled, RTS/DSR set
147 gbGDB_SerialSetup = 1;
149 while( (inb(GDB_SERIAL_PORT + 5) & 0x20) == 0 );
150 outb(GDB_SERIAL_PORT, ch);
153 int getDebugChar(void)
155 if(!gbGDB_SerialSetup) {
156 outb(GDB_SERIAL_PORT + 1, 0x00); // Disable all interrupts
157 outb(GDB_SERIAL_PORT + 3, 0x80); // Enable DLAB (set baud rate divisor)
158 outb(GDB_SERIAL_PORT + 0, 0x0C); // Set divisor to 12 (lo byte) 9600 baud
159 outb(GDB_SERIAL_PORT + 1, 0x00); // (hi byte)
160 outb(GDB_SERIAL_PORT + 3, 0x03); // 8 bits, no parity, one stop bit
161 outb(GDB_SERIAL_PORT + 2, 0xC7); // Enable FIFO with 14-byte threshold and clear it
162 outb(GDB_SERIAL_PORT + 4, 0x0B); // IRQs enabled, RTS/DSR set
163 gbGDB_SerialSetup = 1;
165 while( (inb(GDB_SERIAL_PORT + 5) & 1) == 0) ;
166 return inb(GDB_SERIAL_PORT);
168 #endif /* USE_GDB_STUB */
170 void Debug_PutCharDebug(char ch)
173 __asm__ __volatile__ ( "outb %%al, $0xe9" :: "a"(((Uint8)ch)) );
177 if(!gbDebug_SerialSetup) {
178 outb(SERIAL_PORT + 1, 0x00); // Disable all interrupts
179 outb(SERIAL_PORT + 3, 0x80); // Enable DLAB (set baud rate divisor)
180 outb(SERIAL_PORT + 0, 0x0C); // Set divisor to 12 (lo byte) 9600 baud
181 outb(SERIAL_PORT + 1, 0x00); // (hi byte)
182 outb(SERIAL_PORT + 3, 0x03); // 8 bits, no parity, one stop bit
183 outb(SERIAL_PORT + 2, 0xC7); // Enable FIFO with 14-byte threshold and clear it
184 outb(SERIAL_PORT + 4, 0x0B); // IRQs enabled, RTS/DSR set
185 gbDebug_SerialSetup = 1;
187 while( (inb(SERIAL_PORT + 5) & 0x20) == 0 );
188 outb(SERIAL_PORT, ch);
192 void Debug_PutStringDebug(const char *String)
195 Debug_PutCharDebug(*String++);
198 // === IO Commands ===
199 void outb(Uint16 Port, Uint8 Data)
201 __asm__ __volatile__ ("outb %%al, %%dx"::"d"(Port),"a"(Data));
203 void outw(Uint16 Port, Uint16 Data)
205 __asm__ __volatile__ ("outw %%ax, %%dx"::"d"(Port),"a"(Data));
207 void outd(Uint16 Port, Uint32 Data)
209 __asm__ __volatile__ ("outl %%eax, %%dx"::"d"(Port),"a"(Data));
211 Uint8 inb(Uint16 Port)
214 __asm__ __volatile__ ("inb %%dx, %%al":"=a"(ret):"d"(Port));
217 Uint16 inw(Uint16 Port)
220 __asm__ __volatile__ ("inw %%dx, %%ax":"=a"(ret):"d"(Port));
223 Uint32 ind(Uint16 Port)
226 __asm__ __volatile__ ("inl %%dx, %%eax":"=a"(ret):"d"(Port));
231 * \fn void *memset(void *Dest, int Val, size_t Num)
232 * \brief Do a byte granuality set of Dest
234 void *memset(void *Dest, int Val, size_t Num)
236 Uint32 val = Val&0xFF;
239 __asm__ __volatile__ (
243 :: "D" (Dest), "a" (val), "c" (Num/4), "r" (Num&3));
247 * \brief Set double words
249 void *memsetd(void *Dest, Uint32 Val, size_t Num)
251 __asm__ __volatile__ ("rep stosl" :: "D" (Dest), "a" (Val), "c" (Num));
256 * \fn int memcmp(const void *m1, const void *m2, size_t Num)
257 * \brief Compare two pieces of memory
259 int memcmp(const void *m1, const void *m2, size_t Num)
261 const Uint8 *d1 = m1;
262 const Uint8 *d2 = m2;
263 if( Num == 0 ) return 0; // No bytes are always identical
276 * \fn void *memcpy(void *Dest, const void *Src, size_t Num)
277 * \brief Copy \a Num bytes from \a Src to \a Dest
279 void *memcpy(void *Dest, const void *Src, size_t Num)
281 tVAddr dst = (tVAddr)Dest;
282 tVAddr src = (tVAddr)Src;
283 if( (dst & 3) != (src & 3) )
285 __asm__ __volatile__ ("rep movsb" :: "D" (dst), "S" (src), "c" (Num));
286 // Debug("\nmemcpy:Num=0x%x by %p (UA)", Num, __builtin_return_address(0));
289 else if( Num > 128 && (dst & 15) == (src & 15) )
291 char tmp[16+15]; // Note, this is a hack to save/restor xmm0
292 int count = 16 - (dst & 15);
293 // Debug("\nmemcpy:Num=0x%x by %p (SSE)", Num, __builtin_return_address(0));
297 __asm__ __volatile__ ("rep movsb" : "=D"(dst),"=S"(src): "0"(dst), "1"(src), "c"(count));
301 __asm__ __volatile__ (
302 "movdqa 0(%5), %%xmm0;\n\t"
304 "movdqa 0(%1), %%xmm0;\n\t"
305 "movdqa %%xmm0, 0(%0);\n\t"
309 "movdqa %%xmm0, 0(%5);\n\t"
310 : "=r"(dst),"=r"(src)
311 : "0"(dst), "1"(src), "c"(count), "r" (((tVAddr)tmp+15)&~15)
316 __asm__ __volatile__ ("rep movsb" :: "D"(dst), "S"(src), "c"(count));
321 // Debug("\nmemcpy:Num=0x%x by %p", Num, __builtin_return_address(0));
322 __asm__ __volatile__ (
326 :: "D" (Dest), "S" (Src), "c" (Num/4), "r" (Num&3));
332 * \fn void *memcpyd(void *Dest, const void *Src, size_t Num)
333 * \brief Copy \a Num DWORDs from \a Src to \a Dest
335 void *memcpyd(void *Dest, const void *Src, size_t Num)
337 __asm__ __volatile__ ("rep movsl" :: "D" (Dest), "S" (Src), "c" (Num));
341 #include "../helpers.h"
345 Uint64 DivMod64U(Uint64 Num, Uint64 Div, Uint64 *Rem)
347 if( Div < 0x100000000ULL && Num < 0xFFFFFFFF * Div ) {
349 __asm__ __volatile__(
351 : "=a" (ret_32), "=d" (rem)
352 : "a" ( (Uint32)(Num & 0xFFFFFFFF) ), "d" ((Uint32)(Num >> 32)), "r" (Div)
358 return __divmod64(Num, Div, Rem);
362 * \fn Uint64 __udivdi3(Uint64 Num, Uint64 Den)
363 * \brief Divide two 64-bit integers
365 Uint64 __udivdi3(Uint64 Num, Uint64 Den)
368 __asm__ __volatile__ ("int $0x0");
372 if(Num <= 0xFFFFFFFF && Den <= 0xFFFFFFFF)
373 return (Uint32)Num / (Uint32)Den;
374 if(Den == 1) return Num;
375 if(Den == 2) return Num >> 1; // Speed Hacks
376 if(Den == 4) return Num >> 2; // Speed Hacks
377 if(Den == 8) return Num >> 3; // Speed Hacks
378 if(Den == 16) return Num >> 4; // Speed Hacks
379 if(Den == 32) return Num >> 5; // Speed Hacks
380 if(Den == 1024) return Num >> 10; // Speed Hacks
381 if(Den == 2048) return Num >> 11; // Speed Hacks
382 if(Den == 4096) return Num >> 12;
383 if(Num < Den) return 0;
384 if(Num < Den*2) return 1;
385 if(Num == Den*2) return 2;
387 return __divmod64(Num, Den, NULL);
391 * \fn Uint64 __umoddi3(Uint64 Num, Uint64 Den)
392 * \brief Get the modulus of two 64-bit integers
394 Uint64 __umoddi3(Uint64 Num, Uint64 Den)
398 __asm__ __volatile__ ("int $0x0"); // Call Div by Zero Error
401 if(Den == 1) return 0; // Speed Hacks
402 if(Den == 2) return Num & 1; // Speed Hacks
403 if(Den == 4) return Num & 3; // Speed Hacks
404 if(Den == 8) return Num & 7; // Speed Hacks
405 if(Den == 16) return Num & 15; // Speed Hacks
406 if(Den == 32) return Num & 31; // Speed Hacks
407 if(Den == 1024) return Num & 1023; // Speed Hacks
408 if(Den == 2048) return Num & 2047; // Speed Hacks
409 if(Den == 4096) return Num & 4095; // Speed Hacks
411 if(Num >> 32 == 0 && Den >> 32 == 0)
412 return (Uint32)Num % (Uint32)Den;
414 __divmod64(Num, Den, &ret);
420 EXPORT(memcpy); EXPORT(memset);
422 //EXPORT(memcpyw); EXPORT(memsetw);
423 EXPORT(memcpyd); EXPORT(memsetd);
424 EXPORT(inb); EXPORT(inw); EXPORT(ind);
425 EXPORT(outb); EXPORT(outw); EXPORT(outd);
426 EXPORT(__udivdi3); EXPORT(__umoddi3);