diff --git a/Makefile.target b/Makefile.target index 10f3cb23ac..1767dc1f26 100644 --- a/Makefile.target +++ b/Makefile.target @@ -594,7 +594,7 @@ OBJS+= pxa2xx_lcd.o pxa2xx_mmci.o pxa2xx_pcmcia.o pxa2xx_keypad.o OBJS+= pflash_cfi01.o gumstix.o OBJS+= zaurus.o ide.o serial.o nand.o ecc.o spitz.o tosa.o tc6393xb.o OBJS+= omap1.o omap_lcdc.o omap_dma.o omap_clk.o omap_mmc.o omap_i2c.o -OBJS+= omap2.o omap_dss.o +OBJS+= omap2.o omap_dss.o soc_dma.o OBJS+= palm.o tsc210x.o OBJS+= nseries.o blizzard.o onenand.o vga.o cbus.o tusb6010.o usb-musb.o OBJS+= tsc2005.o diff --git a/hw/omap.h b/hw/omap.h index 7270e202b6..7006ecbe8e 100644 --- a/hw/omap.h +++ b/hw/omap.h @@ -417,14 +417,14 @@ enum omap_dma_model { omap_dma_4, }; -struct omap_dma_s; -struct omap_dma_s *omap_dma_init(target_phys_addr_t base, qemu_irq *irqs, +struct soc_dma_s; +struct soc_dma_s *omap_dma_init(target_phys_addr_t base, qemu_irq *irqs, qemu_irq lcd_irq, struct omap_mpu_state_s *mpu, omap_clk clk, enum omap_dma_model model); -struct omap_dma_s *omap_dma4_init(target_phys_addr_t base, qemu_irq *irqs, +struct soc_dma_s *omap_dma4_init(target_phys_addr_t base, qemu_irq *irqs, struct omap_mpu_state_s *mpu, int fifo, int chans, omap_clk iclk, omap_clk fclk); -void omap_dma_reset(struct omap_dma_s *s); +void omap_dma_reset(struct soc_dma_s *s); struct dma_irq_map { int ih; @@ -494,7 +494,7 @@ struct omap_dma_lcd_channel_s { ram_addr_t phys_framebuffer[2]; qemu_irq irq; struct omap_mpu_state_s *mpu; -} *omap_dma_get_lcdch(struct omap_dma_s *s); +} *omap_dma_get_lcdch(struct soc_dma_s *s); /* * DMA request numbers for OMAP1 @@ -882,7 +882,7 @@ struct omap_mpu_state_s { /* MPU private TIPB peripherals */ struct omap_intr_handler_s *ih[2]; - struct omap_dma_s *dma; + struct soc_dma_s *dma; struct omap_mpu_timer_s *timer[3]; struct omap_watchdog_timer_s *wdt; diff --git a/hw/omap1.c b/hw/omap1.c index 7dab6c8e4b..5854500764 100644 --- a/hw/omap1.c +++ b/hw/omap1.c @@ -24,6 +24,7 @@ #include "sysemu.h" #include "qemu-timer.h" #include "qemu-char.h" +#include "soc_dma.h" /* We use pc-style serial ports. */ #include "pc.h" @@ -4704,6 +4705,12 @@ struct omap_mpu_state_s *omap310_mpu_init(unsigned long sdram_size, s->port[local ].addr_valid = omap_validate_local_addr; s->port[tipb_mpui].addr_valid = omap_validate_tipb_mpui_addr; + /* Register SDRAM and SRAM DMA ports for fast transfers. */ + soc_dma_port_add_mem_ram(s->dma, + emiff_base, OMAP_EMIFF_BASE, s->sdram_size); + soc_dma_port_add_mem_ram(s->dma, + imif_base, OMAP_IMIF_BASE, s->sram_size); + s->timer[0] = omap_mpu_timer_init(0xfffec500, s->irq[0][OMAP_INT_TIMER1], omap_findclk(s, "mputim_ck")); diff --git a/hw/omap2.c b/hw/omap2.c index 8c9ac6a81e..fa7b35c07f 100644 --- a/hw/omap2.c +++ b/hw/omap2.c @@ -26,6 +26,7 @@ #include "qemu-timer.h" #include "qemu-char.h" #include "flash.h" +#include "soc_dma.h" #include "audio/audio.h" /* GP timers */ @@ -4493,6 +4494,10 @@ struct omap_mpu_state_s *omap2420_mpu_init(unsigned long sdram_size, omap_findclk(s, "sdma_fclk")); s->port->addr_valid = omap2_validate_addr; + /* Register SDRAM and SRAM ports for fast DMA transfers. */ + soc_dma_port_add_mem_ram(s->dma, q2_base, OMAP2_Q2_BASE, s->sdram_size); + soc_dma_port_add_mem_ram(s->dma, sram_base, OMAP2_SRAM_BASE, s->sram_size); + s->uart[0] = omap2_uart_init(omap_l4ta(s->l4, 19), s->irq[0][OMAP_INT_24XX_UART1_IRQ], omap_findclk(s, "uart1_fclk"), diff --git a/hw/omap_dma.c b/hw/omap_dma.c index 6c0bd8267f..df4fb01c99 100644 --- a/hw/omap_dma.c +++ b/hw/omap_dma.c @@ -23,6 +23,7 @@ #include "qemu-timer.h" #include "omap.h" #include "irq.h" +#include "soc_dma.h" struct omap_dma_channel_s { /* transfer data */ @@ -66,6 +67,7 @@ struct omap_dma_channel_s { int pending_request; int waiting_end_prog; uint16_t cpc; + int set_update; /* sync type */ int fs; @@ -89,6 +91,8 @@ struct omap_dma_channel_s { int pck_elements; } active_set; + struct soc_dma_ch_s *dma; + /* unused parameters */ int write_mode; int priority; @@ -99,12 +103,11 @@ struct omap_dma_channel_s { }; struct omap_dma_s { - QEMUTimer *tm; + struct soc_dma_s *dma; + struct omap_mpu_state_s *mpu; target_phys_addr_t base; omap_clk clk; - int64_t delay; - uint64_t drq; qemu_irq irq[4]; void (*intr_update)(struct omap_dma_s *s); enum omap_dma_model model; @@ -115,7 +118,6 @@ struct omap_dma_s { uint32_t caps[5]; uint32_t irqen[4]; uint32_t irqstat[4]; - int run_count; int chans; struct omap_dma_channel_s ch[32]; @@ -139,11 +141,10 @@ static inline void omap_dma_interrupts_update(struct omap_dma_s *s) return s->intr_update(s); } -static void omap_dma_channel_load(struct omap_dma_s *s, - struct omap_dma_channel_s *ch) +static void omap_dma_channel_load(struct omap_dma_channel_s *ch) { struct omap_dma_reg_set_s *a = &ch->active_set; - int i; + int i, normal; int omap_3_1 = !ch->omap_3_1_compatible_disable; /* @@ -189,20 +190,50 @@ static void omap_dma_channel_load(struct omap_dma_s *s, default: break; } + + normal = !ch->transparent_copy && !ch->constant_fill && + /* FIFO is big-endian so either (ch->endian[n] == 1) OR + * (ch->endian_lock[n] == 1) mean no endianism conversion. */ + (ch->endian[0] | ch->endian_lock[0]) == + (ch->endian[1] | ch->endian_lock[1]); + for (i = 0; i < 2; i ++) { + /* TODO: for a->frame_delta[i] > 0 still use the fast path, just + * limit min_elems in omap_dma_transfer_setup to the nearest frame + * end. */ + if (!a->elem_delta[i] && normal && + (a->frames == 1 || !a->frame_delta[i])) + ch->dma->type[i] = soc_dma_access_const; + else if (a->elem_delta[i] == ch->data_type && normal && + (a->frames == 1 || !a->frame_delta[i])) + ch->dma->type[i] = soc_dma_access_linear; + else + ch->dma->type[i] = soc_dma_access_other; + + ch->dma->vaddr[i] = ch->addr[i]; + } + soc_dma_ch_update(ch->dma); } static void omap_dma_activate_channel(struct omap_dma_s *s, struct omap_dma_channel_s *ch) { if (!ch->active) { + if (ch->set_update) { + /* It's not clear when the active set is supposed to be + * loaded from registers. We're already loading it when the + * channel is enabled, and for some guests this is not enough + * but that may be also because of a race condition (no + * delays in qemu) in the guest code, which we're just + * working around here. */ + omap_dma_channel_load(ch); + ch->set_update = 0; + } + ch->active = 1; + soc_dma_set_request(ch->dma, 1); if (ch->sync) ch->status |= SYNC; - s->run_count ++; } - - if (s->delay && !qemu_timer_pending(s->tm)) - qemu_mod_timer(s->tm, qemu_get_clock(vm_clock) + s->delay); } static void omap_dma_deactivate_channel(struct omap_dma_s *s, @@ -219,17 +250,14 @@ static void omap_dma_deactivate_channel(struct omap_dma_s *s, /* Don't deactive the channel if it is synchronized and the DMA request is active */ - if (ch->sync && ch->enable && (s->drq & (1 << ch->sync))) + if (ch->sync && ch->enable && (s->dma->drqbmp & (1 << ch->sync))) return; if (ch->active) { ch->active = 0; ch->status &= ~SYNC; - s->run_count --; + soc_dma_set_request(ch->dma, 0); } - - if (!s->run_count) - qemu_del_timer(s->tm); } static void omap_dma_enable_channel(struct omap_dma_s *s, @@ -238,11 +266,11 @@ static void omap_dma_enable_channel(struct omap_dma_s *s, if (!ch->enable) { ch->enable = 1; ch->waiting_end_prog = 0; - omap_dma_channel_load(s, ch); + omap_dma_channel_load(ch); /* TODO: theoretically if ch->sync && ch->prefetch && - * !s->drq[ch->sync], we should also activate and fetch from source - * and then stall until signalled. */ - if ((!ch->sync) || (s->drq & (1 << ch->sync))) + * !s->dma->drqbmp[ch->sync], we should also activate and fetch + * from source and then stall until signalled. */ + if ((!ch->sync) || (s->dma->drqbmp & (1 << ch->sync))) omap_dma_activate_channel(s, ch); } } @@ -338,140 +366,319 @@ static void omap_dma_process_request(struct omap_dma_s *s, int request) omap_dma_interrupts_update(s); } -static void omap_dma_channel_run(struct omap_dma_s *s) +static void omap_dma_transfer_generic(struct soc_dma_ch_s *dma) { - int n = s->chans; - uint16_t status; uint8_t value[4]; - struct omap_dma_port_if_s *src_p, *dest_p; - struct omap_dma_reg_set_s *a; - struct omap_dma_channel_s *ch; - - for (ch = s->ch; n; n --, ch ++) { - if (!ch->active) - continue; - - a = &ch->active_set; - - src_p = &s->mpu->port[ch->port[0]]; - dest_p = &s->mpu->port[ch->port[1]]; - if ((!ch->constant_fill && !src_p->addr_valid(s->mpu, a->src)) || - (!dest_p->addr_valid(s->mpu, a->dest))) { -#if 0 - /* Bus time-out */ - if (ch->interrupts & TIMEOUT_INTR) - ch->status |= TIMEOUT_INTR; - omap_dma_deactivate_channel(s, ch); - continue; + struct omap_dma_channel_s *ch = dma->opaque; + struct omap_dma_reg_set_s *a = &ch->active_set; + int bytes = dma->bytes; +#ifdef MULTI_REQ + uint16_t status = ch->status; #endif - printf("%s: Bus time-out in DMA%i operation\n", - __FUNCTION__, s->chans - n); + + do { + /* Transfer a single element */ + /* FIXME: check the endianness */ + if (!ch->constant_fill) + cpu_physical_memory_read(a->src, value, ch->data_type); + else + *(uint32_t *) value = ch->color; + + if (!ch->transparent_copy || *(uint32_t *) value != ch->color) + cpu_physical_memory_write(a->dest, value, ch->data_type); + + a->src += a->elem_delta[0]; + a->dest += a->elem_delta[1]; + a->element ++; + +#ifndef MULTI_REQ + if (a->element == a->elements) { + /* End of Frame */ + a->element = 0; + a->src += a->frame_delta[0]; + a->dest += a->frame_delta[1]; + a->frame ++; + + /* If the channel is async, update cpc */ + if (!ch->sync) + ch->cpc = a->dest & 0xffff; + } + } while ((bytes -= ch->data_type)); +#else + /* If the channel is element synchronized, deactivate it */ + if (ch->sync && !ch->fs && !ch->bs) + omap_dma_deactivate_channel(s, ch); + + /* If it is the last frame, set the LAST_FRAME interrupt */ + if (a->element == 1 && a->frame == a->frames - 1) + if (ch->interrupts & LAST_FRAME_INTR) + ch->status |= LAST_FRAME_INTR; + + /* If the half of the frame was reached, set the HALF_FRAME + interrupt */ + if (a->element == (a->elements >> 1)) + if (ch->interrupts & HALF_FRAME_INTR) + ch->status |= HALF_FRAME_INTR; + + if (ch->fs && ch->bs) { + a->pck_element ++; + /* Check if a full packet has beed transferred. */ + if (a->pck_element == a->pck_elements) { + a->pck_element = 0; + + /* Set the END_PKT interrupt */ + if ((ch->interrupts & END_PKT_INTR) && !ch->src_sync) + ch->status |= END_PKT_INTR; + + /* If the channel is packet-synchronized, deactivate it */ + if (ch->sync) + omap_dma_deactivate_channel(s, ch); + } } - status = ch->status; - while (status == ch->status && ch->active) { - /* Transfer a single element */ - /* FIXME: check the endianness */ - if (!ch->constant_fill) - cpu_physical_memory_read(a->src, value, ch->data_type); - else - *(uint32_t *) value = ch->color; + if (a->element == a->elements) { + /* End of Frame */ + a->element = 0; + a->src += a->frame_delta[0]; + a->dest += a->frame_delta[1]; + a->frame ++; - if (!ch->transparent_copy || - *(uint32_t *) value != ch->color) - cpu_physical_memory_write(a->dest, value, ch->data_type); - - a->src += a->elem_delta[0]; - a->dest += a->elem_delta[1]; - a->element ++; - - /* If the channel is element synchronized, deactivate it */ - if (ch->sync && !ch->fs && !ch->bs) + /* If the channel is frame synchronized, deactivate it */ + if (ch->sync && ch->fs && !ch->bs) omap_dma_deactivate_channel(s, ch); - /* If it is the last frame, set the LAST_FRAME interrupt */ - if (a->element == 1 && a->frame == a->frames - 1) - if (ch->interrupts & LAST_FRAME_INTR) - ch->status |= LAST_FRAME_INTR; + /* If the channel is async, update cpc */ + if (!ch->sync) + ch->cpc = a->dest & 0xffff; - /* If the half of the frame was reached, set the HALF_FRAME - interrupt */ - if (a->element == (a->elements >> 1)) - if (ch->interrupts & HALF_FRAME_INTR) - ch->status |= HALF_FRAME_INTR; + /* Set the END_FRAME interrupt */ + if (ch->interrupts & END_FRAME_INTR) + ch->status |= END_FRAME_INTR; - if (ch->fs && ch->bs) { - a->pck_element ++; - /* Check if a full packet has beed transferred. */ - if (a->pck_element == a->pck_elements) { - a->pck_element = 0; + if (a->frame == a->frames) { + /* End of Block */ + /* Disable the channel */ - /* Set the END_PKT interrupt */ - if ((ch->interrupts & END_PKT_INTR) && !ch->src_sync) - ch->status |= END_PKT_INTR; - - /* If the channel is packet-synchronized, deactivate it */ - if (ch->sync) - omap_dma_deactivate_channel(s, ch); - } - } - - if (a->element == a->elements) { - /* End of Frame */ - a->element = 0; - a->src += a->frame_delta[0]; - a->dest += a->frame_delta[1]; - a->frame ++; - - /* If the channel is frame synchronized, deactivate it */ - if (ch->sync && ch->fs && !ch->bs) - omap_dma_deactivate_channel(s, ch); - - /* If the channel is async, update cpc */ - if (!ch->sync) - ch->cpc = a->dest & 0xffff; - - /* Set the END_FRAME interrupt */ - if (ch->interrupts & END_FRAME_INTR) - ch->status |= END_FRAME_INTR; - - if (a->frame == a->frames) { - /* End of Block */ - /* Disable the channel */ - - if (ch->omap_3_1_compatible_disable) { + if (ch->omap_3_1_compatible_disable) { + omap_dma_disable_channel(s, ch); + if (ch->link_enabled) + omap_dma_enable_channel(s, + &s->ch[ch->link_next_ch]); + } else { + if (!ch->auto_init) omap_dma_disable_channel(s, ch); - if (ch->link_enabled) - omap_dma_enable_channel(s, - &s->ch[ch->link_next_ch]); - } else { - if (!ch->auto_init) - omap_dma_disable_channel(s, ch); - else if (ch->repeat || ch->end_prog) - omap_dma_channel_load(s, ch); - else { - ch->waiting_end_prog = 1; - omap_dma_deactivate_channel(s, ch); - } + else if (ch->repeat || ch->end_prog) + omap_dma_channel_load(ch); + else { + ch->waiting_end_prog = 1; + omap_dma_deactivate_channel(s, ch); } - - if (ch->interrupts & END_BLOCK_INTR) - ch->status |= END_BLOCK_INTR; } + + if (ch->interrupts & END_BLOCK_INTR) + ch->status |= END_BLOCK_INTR; } } + } while (status == ch->status && ch->active); + + omap_dma_interrupts_update(s); +#endif +} + +enum { + omap_dma_intr_element_sync, + omap_dma_intr_last_frame, + omap_dma_intr_half_frame, + omap_dma_intr_frame, + omap_dma_intr_frame_sync, + omap_dma_intr_packet, + omap_dma_intr_packet_sync, + omap_dma_intr_block, + __omap_dma_intr_last, +}; + +static void omap_dma_transfer_setup(struct soc_dma_ch_s *dma) +{ + struct omap_dma_port_if_s *src_p, *dest_p; + struct omap_dma_reg_set_s *a; + struct omap_dma_channel_s *ch = dma->opaque; + struct omap_dma_s *s = dma->dma->opaque; + int frames, min_elems, elements[__omap_dma_intr_last]; + + a = &ch->active_set; + + src_p = &s->mpu->port[ch->port[0]]; + dest_p = &s->mpu->port[ch->port[1]]; + if ((!ch->constant_fill && !src_p->addr_valid(s->mpu, a->src)) || + (!dest_p->addr_valid(s->mpu, a->dest))) { +#if 0 + /* Bus time-out */ + if (ch->interrupts & TIMEOUT_INTR) + ch->status |= TIMEOUT_INTR; + omap_dma_deactivate_channel(s, ch); + continue; +#endif + printf("%s: Bus time-out in DMA%i operation\n", + __FUNCTION__, dma->num); + } + + min_elems = INT_MAX; + + /* Check all the conditions that terminate the transfer starting + * with those that can occur the soonest. */ +#define INTR_CHECK(cond, id, nelements) \ + if (cond) { \ + elements[id] = nelements; \ + if (elements[id] < min_elems) \ + min_elems = elements[id]; \ + } else \ + elements[id] = INT_MAX; + + /* Elements */ + INTR_CHECK( + ch->sync && !ch->fs && !ch->bs, + omap_dma_intr_element_sync, + 1) + + /* Frames */ + /* TODO: for transfers where entire frames can be read and written + * using memcpy() but a->frame_delta is non-zero, try to still do + * transfers using soc_dma but limit min_elems to a->elements - ... + * See also the TODO in omap_dma_channel_load. */ + INTR_CHECK( + (ch->interrupts & LAST_FRAME_INTR) && + ((a->frame < a->frames - 1) || !a->element), + omap_dma_intr_last_frame, + (a->frames - a->frame - 2) * a->elements + + (a->elements - a->element + 1)) + INTR_CHECK( + ch->interrupts & HALF_FRAME_INTR, + omap_dma_intr_half_frame, + (a->elements >> 1) + + (a->element >= (a->elements >> 1) ? a->elements : 0) - + a->element) + INTR_CHECK( + ch->sync && ch->fs && (ch->interrupts & END_FRAME_INTR), + omap_dma_intr_frame, + a->elements - a->element) + INTR_CHECK( + ch->sync && ch->fs && !ch->bs, + omap_dma_intr_frame_sync, + a->elements - a->element) + + /* Packets */ + INTR_CHECK( + ch->fs && ch->bs && + (ch->interrupts & END_PKT_INTR) && !ch->src_sync, + omap_dma_intr_packet, + a->pck_elements - a->pck_element) + INTR_CHECK( + ch->fs && ch->bs && ch->sync, + omap_dma_intr_packet_sync, + a->pck_elements - a->pck_element) + + /* Blocks */ + INTR_CHECK( + 1, + omap_dma_intr_block, + (a->frames - a->frame - 1) * a->elements + + (a->elements - a->element)) + + dma->bytes = min_elems * ch->data_type; + + /* Set appropriate interrupts and/or deactivate channels */ + +#ifdef MULTI_REQ + /* TODO: should all of this only be done if dma->update, and otherwise + * inside omap_dma_transfer_generic below - check what's faster. */ + if (dma->update) { +#endif + + /* If the channel is element synchronized, deactivate it */ + if (min_elems == elements[omap_dma_intr_element_sync]) + omap_dma_deactivate_channel(s, ch); + + /* If it is the last frame, set the LAST_FRAME interrupt */ + if (min_elems == elements[omap_dma_intr_last_frame]) + ch->status |= LAST_FRAME_INTR; + + /* If exactly half of the frame was reached, set the HALF_FRAME + interrupt */ + if (min_elems == elements[omap_dma_intr_half_frame]) + ch->status |= HALF_FRAME_INTR; + + /* If a full packet has been transferred, set the END_PKT interrupt */ + if (min_elems == elements[omap_dma_intr_packet]) + ch->status |= END_PKT_INTR; + + /* If the channel is packet-synchronized, deactivate it */ + if (min_elems == elements[omap_dma_intr_packet_sync]) + omap_dma_deactivate_channel(s, ch); + + /* If the channel is frame synchronized, deactivate it */ + if (min_elems == elements[omap_dma_intr_frame_sync]) + omap_dma_deactivate_channel(s, ch); + + /* Set the END_FRAME interrupt */ + if (min_elems == elements[omap_dma_intr_frame]) + ch->status |= END_FRAME_INTR; + + if (min_elems == elements[omap_dma_intr_block]) { + /* End of Block */ + /* Disable the channel */ + + if (ch->omap_3_1_compatible_disable) { + omap_dma_disable_channel(s, ch); + if (ch->link_enabled) + omap_dma_enable_channel(s, &s->ch[ch->link_next_ch]); + } else { + if (!ch->auto_init) + omap_dma_disable_channel(s, ch); + else if (ch->repeat || ch->end_prog) + omap_dma_channel_load(ch); + else { + ch->waiting_end_prog = 1; + omap_dma_deactivate_channel(s, ch); + } + } + + if (ch->interrupts & END_BLOCK_INTR) + ch->status |= END_BLOCK_INTR; + } + + /* Update packet number */ + if (ch->fs && ch->bs) { + a->pck_element += min_elems; + a->pck_element %= a->pck_elements; + } + + /* TODO: check if we really need to update anything here or perhaps we + * can skip part of this. */ +#ifndef MULTI_REQ + if (dma->update) { +#endif + a->element += min_elems; + + frames = a->element / a->elements; + a->element = a->element % a->elements; + a->frame += frames; + a->src += min_elems * a->elem_delta[0] + frames * a->frame_delta[0]; + a->dest += min_elems * a->elem_delta[1] + frames * a->frame_delta[1]; + + /* If the channel is async, update cpc */ + if (!ch->sync && frames) + ch->cpc = a->dest & 0xffff; } omap_dma_interrupts_update(s); - if (s->run_count && s->delay) - qemu_mod_timer(s->tm, qemu_get_clock(vm_clock) + s->delay); } -void omap_dma_reset(struct omap_dma_s *s) +void omap_dma_reset(struct soc_dma_s *dma) { int i; + struct omap_dma_s *s = dma->opaque; - qemu_del_timer(s->tm); + soc_dma_reset(s->dma); if (s->model < omap_dma_4) s->gcr = 0x0004; else @@ -479,8 +686,6 @@ void omap_dma_reset(struct omap_dma_s *s) s->ocp = 0x00000000; memset(&s->irqstat, 0, sizeof(s->irqstat)); memset(&s->irqen, 0, sizeof(s->irqen)); - s->drq = 0x00000000; - s->run_count = 0; s->lcd_ch.src = emiff; s->lcd_ch.condition = 0; s->lcd_ch.interrupts = 0; @@ -1161,7 +1366,7 @@ static int omap_dma_sys_write(struct omap_dma_s *s, int offset, uint16_t value) case 0x408: /* DMA_GRST */ if (value & 0x1) - omap_dma_reset(s); + omap_dma_reset(s->dma); break; default: @@ -1338,27 +1543,25 @@ static void omap_dma_request(void *opaque, int drq, int req) struct omap_dma_s *s = (struct omap_dma_s *) opaque; /* The request pins are level triggered in QEMU. */ if (req) { - if (~s->drq & (1 << drq)) { - s->drq |= 1 << drq; + if (~s->dma->drqbmp & (1 << drq)) { + s->dma->drqbmp |= 1 << drq; omap_dma_process_request(s, drq); } } else - s->drq &= ~(1 << drq); + s->dma->drqbmp &= ~(1 << drq); } +/* XXX: this won't be needed once soc_dma knows about clocks. */ static void omap_dma_clk_update(void *opaque, int line, int on) { struct omap_dma_s *s = (struct omap_dma_s *) opaque; + int i; - if (on) { - /* TODO: make a clever calculation */ - s->delay = ticks_per_sec >> 8; - if (s->run_count) - qemu_mod_timer(s->tm, qemu_get_clock(vm_clock) + s->delay); - } else { - s->delay = 0; - qemu_del_timer(s->tm); - } + s->dma->freq = omap_clk_getrate(s->clk); + + for (i = 0; i < s->chans; i ++) + if (s->ch[i].active) + soc_dma_set_request(s->ch[i].dma, on); } static void omap_dma_setcaps(struct omap_dma_s *s) @@ -1407,7 +1610,7 @@ static void omap_dma_setcaps(struct omap_dma_s *s) } } -struct omap_dma_s *omap_dma_init(target_phys_addr_t base, qemu_irq *irqs, +struct soc_dma_s *omap_dma_init(target_phys_addr_t base, qemu_irq *irqs, qemu_irq lcd_irq, struct omap_mpu_state_s *mpu, omap_clk clk, enum omap_dma_model model) { @@ -1428,24 +1631,37 @@ struct omap_dma_s *omap_dma_init(target_phys_addr_t base, qemu_irq *irqs, s->clk = clk; s->lcd_ch.irq = lcd_irq; s->lcd_ch.mpu = mpu; - omap_dma_setcaps(s); + + s->dma = soc_dma_init((model <= omap_dma_3_1) ? 9 : 16); + s->dma->freq = omap_clk_getrate(clk); + s->dma->transfer_fn = omap_dma_transfer_generic; + s->dma->setup_fn = omap_dma_transfer_setup; + s->dma->drq = qemu_allocate_irqs(omap_dma_request, s, 32); + s->dma->opaque = s; + while (num_irqs --) s->ch[num_irqs].irq = irqs[num_irqs]; for (i = 0; i < 3; i ++) { s->ch[i].sibling = &s->ch[i + 6]; s->ch[i + 6].sibling = &s->ch[i]; } - s->tm = qemu_new_timer(vm_clock, (QEMUTimerCB *) omap_dma_channel_run, s); + for (i = (model <= omap_dma_3_1) ? 8 : 15; i >= 0; i --) { + s->ch[i].dma = &s->dma->ch[i]; + s->dma->ch[i].opaque = &s->ch[i]; + } + + omap_dma_setcaps(s); omap_clk_adduser(s->clk, qemu_allocate_irqs(omap_dma_clk_update, s, 1)[0]); - mpu->drq = qemu_allocate_irqs(omap_dma_request, s, 32); - omap_dma_reset(s); + omap_dma_reset(s->dma); omap_dma_clk_update(s, 0, 1); iomemtype = cpu_register_io_memory(0, omap_dma_readfn, omap_dma_writefn, s); cpu_register_physical_memory(s->base, memsize, iomemtype); - return s; + mpu->drq = s->dma->drq; + + return s->dma; } static void omap_dma_interrupts_4_update(struct omap_dma_s *s) @@ -1646,7 +1862,7 @@ static void omap_dma4_write(void *opaque, target_phys_addr_t addr, case 0x2c: /* DMA4_OCP_SYSCONFIG */ if (value & 2) /* SOFTRESET */ - omap_dma_reset(s); + omap_dma_reset(s->dma); s->ocp = value & 0x3321; if (((s->ocp >> 12) & 3) == 3) /* MIDLEMODE */ fprintf(stderr, "%s: invalid DMA power mode\n", __FUNCTION__); @@ -1728,7 +1944,7 @@ static void omap_dma4_write(void *opaque, target_phys_addr_t addr, ch->endian[1] =(value >> 19) & 1; ch->endian_lock[1] =(value >> 18) & 1; if (ch->endian[0] != ch->endian[1]) - fprintf(stderr, "%s: DMA endianned conversion enable attempt\n", + fprintf(stderr, "%s: DMA endiannes conversion enable attempt\n", __FUNCTION__); ch->write_mode = (value >> 16) & 3; ch->burst[1] = (value & 0xc000) >> 14; @@ -1746,35 +1962,43 @@ static void omap_dma4_write(void *opaque, target_phys_addr_t addr, break; case 0x14: /* DMA4_CEN */ + ch->set_update = 1; ch->elements = value & 0xffffff; break; case 0x18: /* DMA4_CFN */ ch->frames = value & 0xffff; + ch->set_update = 1; break; case 0x1c: /* DMA4_CSSA */ ch->addr[0] = (target_phys_addr_t) (uint32_t) value; + ch->set_update = 1; break; case 0x20: /* DMA4_CDSA */ ch->addr[1] = (target_phys_addr_t) (uint32_t) value; + ch->set_update = 1; break; case 0x24: /* DMA4_CSEI */ ch->element_index[0] = (int16_t) value; + ch->set_update = 1; break; case 0x28: /* DMA4_CSFI */ ch->frame_index[0] = (int32_t) value; + ch->set_update = 1; break; case 0x2c: /* DMA4_CDEI */ ch->element_index[1] = (int16_t) value; + ch->set_update = 1; break; case 0x30: /* DMA4_CDFI */ ch->frame_index[1] = (int32_t) value; + ch->set_update = 1; break; case 0x44: /* DMA4_COLOR */ @@ -1806,11 +2030,11 @@ static CPUWriteMemoryFunc *omap_dma4_writefn[] = { omap_dma4_write, }; -struct omap_dma_s *omap_dma4_init(target_phys_addr_t base, qemu_irq *irqs, +struct soc_dma_s *omap_dma4_init(target_phys_addr_t base, qemu_irq *irqs, struct omap_mpu_state_s *mpu, int fifo, int chans, omap_clk iclk, omap_clk fclk) { - int iomemtype; + int iomemtype, i; struct omap_dma_s *s = (struct omap_dma_s *) qemu_mallocz(sizeof(struct omap_dma_s)); @@ -1819,23 +2043,38 @@ struct omap_dma_s *omap_dma4_init(target_phys_addr_t base, qemu_irq *irqs, s->chans = chans; s->mpu = mpu; s->clk = fclk; + + s->dma = soc_dma_init(s->chans); + s->dma->freq = omap_clk_getrate(fclk); + s->dma->transfer_fn = omap_dma_transfer_generic; + s->dma->setup_fn = omap_dma_transfer_setup; + s->dma->drq = qemu_allocate_irqs(omap_dma_request, s, 64); + s->dma->opaque = s; + for (i = 0; i < s->chans; i ++) { + s->ch[i].dma = &s->dma->ch[i]; + s->dma->ch[i].opaque = &s->ch[i]; + } + memcpy(&s->irq, irqs, sizeof(s->irq)); s->intr_update = omap_dma_interrupts_4_update; + omap_dma_setcaps(s); - s->tm = qemu_new_timer(vm_clock, (QEMUTimerCB *) omap_dma_channel_run, s); omap_clk_adduser(s->clk, qemu_allocate_irqs(omap_dma_clk_update, s, 1)[0]); - mpu->drq = qemu_allocate_irqs(omap_dma_request, s, 64); - omap_dma_reset(s); - omap_dma_clk_update(s, 0, 1); + omap_dma_reset(s->dma); + omap_dma_clk_update(s, 0, !!s->dma->freq); iomemtype = cpu_register_io_memory(0, omap_dma4_readfn, omap_dma4_writefn, s); cpu_register_physical_memory(s->base, 0x1000, iomemtype); - return s; + mpu->drq = s->dma->drq; + + return s->dma; } -struct omap_dma_lcd_channel_s *omap_dma_get_lcdch(struct omap_dma_s *s) +struct omap_dma_lcd_channel_s *omap_dma_get_lcdch(struct soc_dma_s *dma) { + struct omap_dma_s *s = dma->opaque; + return &s->lcd_ch; } diff --git a/hw/soc_dma.c b/hw/soc_dma.c new file mode 100644 index 0000000000..f46e260600 --- /dev/null +++ b/hw/soc_dma.c @@ -0,0 +1,366 @@ +/* + * On-chip DMA controller framework. + * + * Copyright (C) 2008 Nokia Corporation + * Written by Andrzej Zaborowski + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License as + * published by the Free Software Foundation; either version 2 or + * (at your option) version 3 of the License. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, + * MA 02111-1307 USA + */ +#include "qemu-common.h" +#include "qemu-timer.h" +#include "soc_dma.h" + +void transfer_mem2mem(struct soc_dma_ch_s *ch) +{ + memcpy(ch->paddr[0], ch->paddr[1], ch->bytes); + ch->paddr[0] += ch->bytes; + ch->paddr[1] += ch->bytes; +} + +void transfer_mem2fifo(struct soc_dma_ch_s *ch) +{ + ch->io_fn[1](ch->io_opaque[1], ch->paddr[0], ch->bytes); + ch->paddr[0] += ch->bytes; +} + +void transfer_fifo2mem(struct soc_dma_ch_s *ch) +{ + ch->io_fn[0](ch->io_opaque[0], ch->paddr[1], ch->bytes); + ch->paddr[1] += ch->bytes; +} + +/* This is further optimisable but isn't very important because often + * DMA peripherals forbid this kind of transfers and even when they don't, + * oprating systems may not need to use them. */ +static void *fifo_buf; +static int fifo_size; +void transfer_fifo2fifo(struct soc_dma_ch_s *ch) +{ + if (ch->bytes < fifo_size) + fifo_buf = realloc(fifo_buf, fifo_size = ch->bytes); + + /* Implement as transfer_fifo2linear + transfer_linear2fifo. */ + ch->io_fn[0](ch->io_opaque[0], fifo_buf, ch->bytes); + ch->io_fn[1](ch->io_opaque[1], fifo_buf, ch->bytes); +} + +struct dma_s { + struct soc_dma_s soc; + int chnum; + uint64_t ch_enable_mask; + int64_t channel_freq; + int enabled_count; + + struct memmap_entry_s { + enum soc_dma_port_type type; + target_phys_addr_t addr; + union { + struct { + void *opaque; + soc_dma_io_t fn; + int out; + } fifo; + struct { + void *base; + size_t size; + } mem; + } u; + } *memmap; + int memmap_size; + + struct soc_dma_ch_s ch[0]; +}; + +static void soc_dma_ch_schedule(struct soc_dma_ch_s *ch, int delay_bytes) +{ + int64_t now = qemu_get_clock(vm_clock); + struct dma_s *dma = (struct dma_s *) ch->dma; + + qemu_mod_timer(ch->timer, now + delay_bytes / dma->channel_freq); +} + +static void soc_dma_ch_run(void *opaque) +{ + struct soc_dma_ch_s *ch = (struct soc_dma_ch_s *) opaque; + + ch->running = 1; + ch->dma->setup_fn(ch); + ch->transfer_fn(ch); + ch->running = 0; + + if (ch->enable) + soc_dma_ch_schedule(ch, ch->bytes); + ch->bytes = 0; +} + +static inline struct memmap_entry_s *soc_dma_lookup(struct dma_s *dma, + target_phys_addr_t addr) +{ + struct memmap_entry_s *lo; + int hi; + + lo = dma->memmap; + hi = dma->memmap_size; + + while (hi > 1) { + hi /= 2; + if (lo[hi].addr <= addr) + lo += hi; + } + + return lo; +} + +static inline enum soc_dma_port_type soc_dma_ch_update_type( + struct soc_dma_ch_s *ch, int port) +{ + struct dma_s *dma = (struct dma_s *) ch->dma; + struct memmap_entry_s *entry = soc_dma_lookup(dma, ch->vaddr[port]); + + if (entry->type == soc_dma_port_fifo) { + while (entry < dma->memmap + dma->memmap_size && + entry->u.fifo.out != port) + entry ++; + if (entry->addr != ch->vaddr[port] || entry->u.fifo.out != port) + return soc_dma_port_other; + + if (ch->type[port] != soc_dma_access_const) + return soc_dma_port_other; + + ch->io_fn[port] = entry->u.fifo.fn; + ch->io_opaque[port] = entry->u.fifo.opaque; + return soc_dma_port_fifo; + } else if (entry->type == soc_dma_port_mem) { + if (entry->addr > ch->vaddr[port] || + entry->addr + entry->u.mem.size <= ch->vaddr[port]) + return soc_dma_port_other; + + /* TODO: support constant memory address for source port as used for + * drawing solid rectangles by PalmOS(R). */ + if (ch->type[port] != soc_dma_access_const) + return soc_dma_port_other; + + ch->paddr[port] = (uint8_t *) entry->u.mem.base + + (ch->vaddr[port] - entry->addr); + /* TODO: save bytes left to the end of the mapping somewhere so we + * can check we're not reading beyond it. */ + return soc_dma_port_mem; + } else + return soc_dma_port_other; +} + +void soc_dma_ch_update(struct soc_dma_ch_s *ch) +{ + enum soc_dma_port_type src, dst; + + src = soc_dma_ch_update_type(ch, 0); + if (src == soc_dma_port_other) { + ch->update = 0; + ch->transfer_fn = ch->dma->transfer_fn; + return; + } + dst = soc_dma_ch_update_type(ch, 1); + + /* TODO: use src and dst as array indices. */ + if (src == soc_dma_port_mem && dst == soc_dma_port_mem) + ch->transfer_fn = transfer_mem2mem; + else if (src == soc_dma_port_mem && dst == soc_dma_port_fifo) + ch->transfer_fn = transfer_mem2fifo; + else if (src == soc_dma_port_fifo && dst == soc_dma_port_mem) + ch->transfer_fn = transfer_fifo2mem; + else if (src == soc_dma_port_fifo && dst == soc_dma_port_fifo) + ch->transfer_fn = transfer_fifo2fifo; + else + ch->transfer_fn = ch->dma->transfer_fn; + + ch->update = (dst != soc_dma_port_other); +} + +static void soc_dma_ch_freq_update(struct dma_s *s) +{ + if (s->enabled_count) + /* We completely ignore channel priorities and stuff */ + s->channel_freq = s->soc.freq / s->enabled_count; + else + /* TODO: Signal that we want to disable the functional clock and let + * the platform code decide what to do with it, i.e. check that + * auto-idle is enabled in the clock controller and if we are stopping + * the clock, do the same with any parent clocks that had only one + * user keeping them on and auto-idle enabled. */; +} + +void soc_dma_set_request(struct soc_dma_ch_s *ch, int level) +{ + struct dma_s *dma = (struct dma_s *) ch->dma; + + dma->enabled_count += level - ch->enable; + + if (level) + dma->ch_enable_mask |= 1 << ch->num; + else + dma->ch_enable_mask &= ~(1 << ch->num); + + if (level != ch->enable) { + soc_dma_ch_freq_update(dma); + ch->enable = level; + + if (!ch->enable) + qemu_del_timer(ch->timer); + else if (!ch->running) + soc_dma_ch_run(ch); + else + soc_dma_ch_schedule(ch, 1); + } +} + +void soc_dma_reset(struct soc_dma_s *soc) +{ + struct dma_s *s = (struct dma_s *) soc; + + s->soc.drqbmp = 0; + s->ch_enable_mask = 0; + s->enabled_count = 0; + soc_dma_ch_freq_update(s); +} + +/* TODO: take a functional-clock argument */ +struct soc_dma_s *soc_dma_init(int n) +{ + int i; + struct dma_s *s = qemu_mallocz(sizeof(*s) + n * sizeof(*s->ch)); + + s->chnum = n; + s->soc.ch = s->ch; + for (i = 0; i < n; i ++) { + s->ch[i].dma = &s->soc; + s->ch[i].num = i; + s->ch[i].timer = qemu_new_timer(vm_clock, soc_dma_ch_run, &s->ch[i]); + } + + soc_dma_reset(&s->soc); + + return &s->soc; +} + +void soc_dma_port_add_fifo(struct soc_dma_s *soc, target_phys_addr_t virt_base, + soc_dma_io_t fn, void *opaque, int out) +{ + struct memmap_entry_s *entry; + struct dma_s *dma = (struct dma_s *) soc; + + dma->memmap = realloc(dma->memmap, sizeof(*entry) * + (dma->memmap_size + 1)); + entry = soc_dma_lookup(dma, virt_base); + + if (dma->memmap_size) { + if (entry->type == soc_dma_port_mem) { + if (entry->addr <= virt_base && + entry->addr + entry->u.mem.size > virt_base) { + fprintf(stderr, "%s: FIFO at " TARGET_FMT_lx + " collides with RAM region at " TARGET_FMT_lx + "-" TARGET_FMT_lx "\n", __FUNCTION__, + (target_ulong) virt_base, + (target_ulong) entry->addr, (target_ulong) + (entry->addr + entry->u.mem.size)); + exit(-1); + } + + if (entry->addr <= virt_base) + entry ++; + } else + while (entry < dma->memmap + dma->memmap_size && + entry->addr <= virt_base) { + if (entry->addr == virt_base && entry->u.fifo.out == out) { + fprintf(stderr, "%s: FIFO at " TARGET_FMT_lx + " collides FIFO at " TARGET_FMT_lx "\n", + __FUNCTION__, (target_ulong) virt_base, + (target_ulong) entry->addr); + exit(-1); + } + + entry ++; + } + + memmove(entry + 1, entry, + (uint8_t *) (dma->memmap + dma->memmap_size ++) - + (uint8_t *) entry); + } else + dma->memmap_size ++; + + entry->addr = virt_base; + entry->type = soc_dma_port_fifo; + entry->u.fifo.fn = fn; + entry->u.fifo.opaque = opaque; + entry->u.fifo.out = out; +} + +void soc_dma_port_add_mem(struct soc_dma_s *soc, uint8_t *phys_base, + target_phys_addr_t virt_base, size_t size) +{ + struct memmap_entry_s *entry; + struct dma_s *dma = (struct dma_s *) soc; + + dma->memmap = realloc(dma->memmap, sizeof(*entry) * + (dma->memmap_size + 1)); + entry = soc_dma_lookup(dma, virt_base); + + if (dma->memmap_size) { + if (entry->type == soc_dma_port_mem) { + if ((entry->addr >= virt_base && entry->addr < virt_base + size) || + (entry->addr <= virt_base && + entry->addr + entry->u.mem.size > virt_base)) { + fprintf(stderr, "%s: RAM at " TARGET_FMT_lx "-" TARGET_FMT_lx + " collides with RAM region at " TARGET_FMT_lx + "-" TARGET_FMT_lx "\n", __FUNCTION__, + (target_ulong) virt_base, + (target_ulong) (virt_base + size), + (target_ulong) entry->addr, (target_ulong) + (entry->addr + entry->u.mem.size)); + exit(-1); + } + + if (entry->addr <= virt_base) + entry ++; + } else { + if (entry->addr >= virt_base && + entry->addr < virt_base + size) { + fprintf(stderr, "%s: RAM at " TARGET_FMT_lx "-" TARGET_FMT_lx + " collides with FIFO at " TARGET_FMT_lx + "\n", __FUNCTION__, + (target_ulong) virt_base, + (target_ulong) (virt_base + size), + (target_ulong) entry->addr); + exit(-1); + } + + while (entry < dma->memmap + dma->memmap_size && + entry->addr <= virt_base) + entry ++; + } + + memmove(entry + 1, entry, + (uint8_t *) (dma->memmap + dma->memmap_size ++) - + (uint8_t *) entry); + } else + dma->memmap_size ++; + + entry->addr = virt_base; + entry->type = soc_dma_port_mem; + entry->u.mem.base = phys_base; + entry->u.mem.size = size; +} + +/* TODO: port removal for ports like PCMCIA memory */ diff --git a/hw/soc_dma.h b/hw/soc_dma.h new file mode 100644 index 0000000000..6ce8b9e579 --- /dev/null +++ b/hw/soc_dma.h @@ -0,0 +1,115 @@ +/* + * On-chip DMA controller framework. + * + * Copyright (C) 2008 Nokia Corporation + * Written by Andrzej Zaborowski + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License as + * published by the Free Software Foundation; either version 2 or + * (at your option) version 3 of the License. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, + * MA 02111-1307 USA + */ + +struct soc_dma_s; +struct soc_dma_ch_s; +typedef void (*soc_dma_io_t)(void *opaque, uint8_t *buf, int len); +typedef void (*soc_dma_transfer_t)(struct soc_dma_ch_s *ch); + +enum soc_dma_port_type { + soc_dma_port_mem, + soc_dma_port_fifo, + soc_dma_port_other, +}; + +enum soc_dma_access_type { + soc_dma_access_const, + soc_dma_access_linear, + soc_dma_access_other, +}; + +struct soc_dma_ch_s { + /* Private */ + struct soc_dma_s *dma; + int num; + QEMUTimer *timer; + + /* Set by soc_dma.c */ + int enable; + int update; + + /* This should be set by dma->setup_fn(). */ + int bytes; + /* Initialised by the DMA module, call soc_dma_ch_update after writing. */ + enum soc_dma_access_type type[2]; + target_phys_addr_t vaddr[2]; /* Updated by .transfer_fn(). */ + /* Private */ + void *paddr[2]; + soc_dma_io_t io_fn[2]; + void *io_opaque[2]; + + int running; + soc_dma_transfer_t transfer_fn; + + /* Set and used by the DMA module. */ + void *opaque; +}; + +struct soc_dma_s { + /* Following fields are set by the SoC DMA module and can be used + * by anybody. */ + uint64_t drqbmp; /* Is zeroed by soc_dma_reset() */ + qemu_irq *drq; + void *opaque; + int64_t freq; + soc_dma_transfer_t transfer_fn; + soc_dma_transfer_t setup_fn; + /* Set by soc_dma_init() for use by the DMA module. */ + struct soc_dma_ch_s *ch; +}; + +/* Call to activate or stop a DMA channel. */ +void soc_dma_set_request(struct soc_dma_ch_s *ch, int level); +/* Call after every write to one of the following fields and before + * calling soc_dma_set_request(ch, 1): + * ch->type[0...1], + * ch->vaddr[0...1], + * ch->paddr[0...1], + * or after a soc_dma_port_add_fifo() or soc_dma_port_add_mem(). */ +void soc_dma_ch_update(struct soc_dma_ch_s *ch); + +/* The SoC should call this when the DMA module is being reset. */ +void soc_dma_reset(struct soc_dma_s *s); +struct soc_dma_s *soc_dma_init(int n); + +void soc_dma_port_add_fifo(struct soc_dma_s *dma, target_phys_addr_t virt_base, + soc_dma_io_t fn, void *opaque, int out); +void soc_dma_port_add_mem(struct soc_dma_s *dma, uint8_t *phys_base, + target_phys_addr_t virt_base, size_t size); + +static inline void soc_dma_port_add_fifo_in(struct soc_dma_s *dma, + target_phys_addr_t virt_base, soc_dma_io_t fn, void *opaque) +{ + return soc_dma_port_add_fifo(dma, virt_base, fn, opaque, 0); +} + +static inline void soc_dma_port_add_fifo_out(struct soc_dma_s *dma, + target_phys_addr_t virt_base, soc_dma_io_t fn, void *opaque) +{ + return soc_dma_port_add_fifo(dma, virt_base, fn, opaque, 1); +} + +static inline void soc_dma_port_add_mem_ram(struct soc_dma_s *dma, + ram_addr_t offset, target_phys_addr_t virt_base, size_t size) +{ + return soc_dma_port_add_mem(dma, phys_ram_base + offset, virt_base, size); +}