Merge remote-tracking branch 'kwolf/for-anthony' into staging

6 files changed, 267 insertions, 29 deletions

diff --git a/hw/ide/core.c b/hw/ide/core.c
index 95beb175b3..ca17a436c0 100644
--- a/hw/ide/core.c
+++ b/hw/ide/core.c
@@ -78,7 +78,7 @@ static void ide_identify(IDEState *s)
 {
     uint16_t *p;
     unsigned int oldsize;
-    IDEDevice *dev;
+    IDEDevice *dev = s->unit ? s->bus->slave : s->bus->master;
 
     if (s->identify_set) {
 	memcpy(s->io_buffer, s->identify_data, sizeof(s->identify_data));
@@ -124,6 +124,9 @@ static void ide_identify(IDEState *s)
     put_le16(p + 66, 120);
     put_le16(p + 67, 120);
     put_le16(p + 68, 120);
+    if (dev && dev->conf.discard_granularity) {
+        put_le16(p + 69, (1 << 14)); /* determinate TRIM behavior */
+    }
 
     if (s->ncq_queues) {
         put_le16(p + 75, s->ncq_queues - 1);
@@ -154,9 +157,12 @@ static void ide_identify(IDEState *s)
     put_le16(p + 101, s->nb_sectors >> 16);
     put_le16(p + 102, s->nb_sectors >> 32);
     put_le16(p + 103, s->nb_sectors >> 48);
-    dev = s->unit ? s->bus->slave : s->bus->master;
+
     if (dev && dev->conf.physical_block_size)
         put_le16(p + 106, 0x6000 | get_physical_block_exp(&dev->conf));
+    if (dev && dev->conf.discard_granularity) {
+        put_le16(p + 169, 1); /* TRIM support */
+    }
 
     memcpy(s->identify_data, p, sizeof(s->identify_data));
     s->identify_set = 1;
@@ -299,6 +305,74 @@ static void ide_set_signature(IDEState *s)
     }
 }
 
+typedef struct TrimAIOCB {
+    BlockDriverAIOCB common;
+    QEMUBH *bh;
+    int ret;
+} TrimAIOCB;
+
+static void trim_aio_cancel(BlockDriverAIOCB *acb)
+{
+    TrimAIOCB *iocb = container_of(acb, TrimAIOCB, common);
+
+    qemu_bh_delete(iocb->bh);
+    iocb->bh = NULL;
+    qemu_aio_release(iocb);
+}
+
+static AIOPool trim_aio_pool = {
+    .aiocb_size         = sizeof(TrimAIOCB),
+    .cancel             = trim_aio_cancel,
+};
+
+static void ide_trim_bh_cb(void *opaque)
+{
+    TrimAIOCB *iocb = opaque;
+
+    iocb->common.cb(iocb->common.opaque, iocb->ret);
+
+    qemu_bh_delete(iocb->bh);
+    iocb->bh = NULL;
+
+    qemu_aio_release(iocb);
+}
+
+BlockDriverAIOCB *ide_issue_trim(BlockDriverState *bs,
+        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
+        BlockDriverCompletionFunc *cb, void *opaque)
+{
+    TrimAIOCB *iocb;
+    int i, j, ret;
+
+    iocb = qemu_aio_get(&trim_aio_pool, bs, cb, opaque);
+    iocb->bh = qemu_bh_new(ide_trim_bh_cb, iocb);
+    iocb->ret = 0;
+
+    for (j = 0; j < qiov->niov; j++) {
+        uint64_t *buffer = qiov->iov[j].iov_base;
+
+        for (i = 0; i < qiov->iov[j].iov_len / 8; i++) {
+            /* 6-byte LBA + 2-byte range per entry */
+            uint64_t entry = le64_to_cpu(buffer[i]);
+            uint64_t sector = entry & 0x0000ffffffffffffULL;
+            uint16_t count = entry >> 48;
+
+            if (count == 0) {
+                break;
+            }
+
+            ret = bdrv_discard(bs, sector, count);
+            if (!iocb->ret) {
+                iocb->ret = ret;
+            }
+        }
+    }
+
+    qemu_bh_schedule(iocb->bh);
+
+    return &iocb->common;
+}
+
 static inline void ide_abort_command(IDEState *s)
 {
     s->status = READY_STAT | ERR_STAT;
@@ -446,7 +520,7 @@ static int ide_handle_rw_error(IDEState *s, int error, int op)
     if ((error == ENOSPC && action == BLOCK_ERR_STOP_ENOSPC)
             || action == BLOCK_ERR_STOP_ANY) {
         s->bus->dma->ops->set_unit(s->bus->dma, s->unit);
-        s->bus->dma->ops->add_status(s->bus->dma, op);
+        s->bus->error_status = op;
         bdrv_mon_event(s->bs, BDRV_ACTION_STOP, is_read);
         vm_stop(VMSTOP_DISKFULL);
     } else {
@@ -472,8 +546,11 @@ handle_rw_error:
     if (ret < 0) {
         int op = BM_STATUS_DMA_RETRY;
 
-        if (s->is_read)
+        if (s->dma_cmd == IDE_DMA_READ)
             op |= BM_STATUS_RETRY_READ;
+        else if (s->dma_cmd == IDE_DMA_TRIM)
+            op |= BM_STATUS_RETRY_TRIM;
+
         if (ide_handle_rw_error(s, -ret, op)) {
             return;
         }
@@ -482,7 +559,7 @@ handle_rw_error:
     n = s->io_buffer_size >> 9;
     sector_num = ide_get_sector(s);
     if (n > 0) {
-        dma_buf_commit(s, s->is_read);
+        dma_buf_commit(s, ide_cmd_is_read(s));
         sector_num += n;
         ide_set_sector(s, sector_num);
         s->nsector -= n;
@@ -499,23 +576,30 @@ handle_rw_error:
     n = s->nsector;
     s->io_buffer_index = 0;
     s->io_buffer_size = n * 512;
-    if (s->bus->dma->ops->prepare_buf(s->bus->dma, s->is_read) == 0) {
+    if (s->bus->dma->ops->prepare_buf(s->bus->dma, ide_cmd_is_read(s)) == 0) {
         /* The PRDs were too short. Reset the Active bit, but don't raise an
          * interrupt. */
         goto eot;
     }
 
 #ifdef DEBUG_AIO
-    printf("ide_dma_cb: sector_num=%" PRId64 " n=%d, is_read=%d\n",
-           sector_num, n, s->is_read);
+    printf("ide_dma_cb: sector_num=%" PRId64 " n=%d, cmd_cmd=%d\n",
+           sector_num, n, s->dma_cmd);
 #endif
 
-    if (s->is_read) {
+    switch (s->dma_cmd) {
+    case IDE_DMA_READ:
         s->bus->dma->aiocb = dma_bdrv_read(s->bs, &s->sg, sector_num,
                                            ide_dma_cb, s);
-    } else {
+        break;
+    case IDE_DMA_WRITE:
         s->bus->dma->aiocb = dma_bdrv_write(s->bs, &s->sg, sector_num,
                                             ide_dma_cb, s);
+        break;
+    case IDE_DMA_TRIM:
+        s->bus->dma->aiocb = dma_bdrv_io(s->bs, &s->sg, sector_num,
+                                         ide_issue_trim, ide_dma_cb, s, 1);
+        break;
     }
 
     if (!s->bus->dma->aiocb) {
@@ -528,12 +612,12 @@ eot:
    ide_set_inactive(s);
 }
 
-static void ide_sector_start_dma(IDEState *s, int is_read)
+static void ide_sector_start_dma(IDEState *s, enum ide_dma_cmd dma_cmd)
 {
     s->status = READY_STAT | SEEK_STAT | DRQ_STAT | BUSY_STAT;
     s->io_buffer_index = 0;
     s->io_buffer_size = 0;
-    s->is_read = is_read;
+    s->dma_cmd = dma_cmd;
     s->bus->dma->ops->start_dma(s->bus->dma, s, ide_dma_cb);
 }
 
@@ -815,6 +899,18 @@ void ide_exec_cmd(IDEBus *bus, uint32_t val)
         return;
 
     switch(val) {
+    case WIN_DSM:
+        switch (s->feature) {
+        case DSM_TRIM:
+            if (!s->bs) {
+                goto abort_cmd;
+            }
+            ide_sector_start_dma(s, IDE_DMA_TRIM);
+            break;
+        default:
+            goto abort_cmd;
+        }
+        break;
     case WIN_IDENTIFY:
         if (s->bs && s->drive_kind != IDE_CD) {
             if (s->drive_kind != IDE_CFATA)
@@ -916,7 +1012,7 @@ void ide_exec_cmd(IDEBus *bus, uint32_t val)
         if (!s->bs)
             goto abort_cmd;
 	ide_cmd_lba48_transform(s, lba48);
-        ide_sector_start_dma(s, 1);
+        ide_sector_start_dma(s, IDE_DMA_READ);
         break;
 	case WIN_WRITEDMA_EXT:
 	lba48 = 1;
@@ -925,7 +1021,7 @@ void ide_exec_cmd(IDEBus *bus, uint32_t val)
         if (!s->bs)
             goto abort_cmd;
 	ide_cmd_lba48_transform(s, lba48);
-        ide_sector_start_dma(s, 0);
+        ide_sector_start_dma(s, IDE_DMA_WRITE);
         s->media_changed = 1;
         break;
     case WIN_READ_NATIVE_MAX_EXT:
@@ -1837,7 +1933,8 @@ static bool ide_drive_pio_state_needed(void *opaque)
 {
     IDEState *s = opaque;
 
-    return (s->status & DRQ_STAT) != 0;
+    return ((s->status & DRQ_STAT) != 0)
+        || (s->bus->error_status & BM_STATUS_PIO_RETRY);
 }
 
 static bool ide_atapi_gesn_needed(void *opaque)
@@ -1847,6 +1944,13 @@ static bool ide_atapi_gesn_needed(void *opaque)
     return s->events.new_media || s->events.eject_request;
 }
 
+static bool ide_error_needed(void *opaque)
+{
+    IDEBus *bus = opaque;
+
+    return (bus->error_status != 0);
+}
+
 /* Fields for GET_EVENT_STATUS_NOTIFICATION ATAPI command */
 const VMStateDescription vmstate_ide_atapi_gesn_state = {
     .name ="ide_drive/atapi/gesn_state",
@@ -1856,6 +1960,7 @@ const VMStateDescription vmstate_ide_atapi_gesn_state = {
     .fields = (VMStateField []) {
         VMSTATE_BOOL(events.new_media, IDEState),
         VMSTATE_BOOL(events.eject_request, IDEState),
+        VMSTATE_END_OF_LIST()
     }
 };
 
@@ -1921,6 +2026,17 @@ const VMStateDescription vmstate_ide_drive = {
     }
 };
 
+const VMStateDescription vmstate_ide_error_status = {
+    .name ="ide_bus/error",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .minimum_version_id_old = 1,
+    .fields = (VMStateField []) {
+        VMSTATE_INT32(error_status, IDEBus),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
 const VMStateDescription vmstate_ide_bus = {
     .name = "ide_bus",
     .version_id = 1,
@@ -1930,6 +2046,14 @@ const VMStateDescription vmstate_ide_bus = {
         VMSTATE_UINT8(cmd, IDEBus),
         VMSTATE_UINT8(unit, IDEBus),
         VMSTATE_END_OF_LIST()
+    },
+    .subsections = (VMStateSubsection []) {
+        {
+            .vmsd = &vmstate_ide_error_status,
+            .needed = ide_error_needed,
+        }, {
+            /* empty */
+        }
     }
 };
 
diff --git a/hw/ide/internal.h b/hw/ide/internal.h
index c2b35ec5e6..02e805f070 100644
--- a/hw/ide/internal.h
+++ b/hw/ide/internal.h
@@ -62,7 +62,11 @@ typedef struct IDEDMAOps IDEDMAOps;
  */
 #define CFA_REQ_EXT_ERROR_CODE		0x03 /* CFA Request Extended Error Code */
 /*
- *	0x04->0x07 Reserved
+ *      0x04->0x05 Reserved
+ */
+#define WIN_DSM                         0x06
+/*
+ *      0x07 Reserved
  */
 #define WIN_SRST			0x08 /* ATAPI soft reset command */
 #define WIN_DEVICE_RESET		0x08
@@ -190,6 +194,9 @@ typedef struct IDEDMAOps IDEDMAOps;
 
 #define IDE_DMA_BUF_SECTORS 256
 
+/* feature values for Data Set Management */
+#define DSM_TRIM                        0x01
+
 #if (IDE_DMA_BUF_SECTORS < MAX_MULT_SECTORS)
 #error "IDE_DMA_BUF_SECTORS must be bigger or equal to MAX_MULT_SECTORS"
 #endif
@@ -379,6 +386,15 @@ struct unreported_events {
     bool new_media;
 };
 
+enum ide_dma_cmd {
+    IDE_DMA_READ,
+    IDE_DMA_WRITE,
+    IDE_DMA_TRIM,
+};
+
+#define ide_cmd_is_read(s) \
+	((s)->dma_cmd == IDE_DMA_READ)
+
 /* NOTE: IDEState represents in fact one drive */
 struct IDEState {
     IDEBus *bus;
@@ -446,7 +462,7 @@ struct IDEState {
     uint32_t mdata_size;
     uint8_t *mdata_storage;
     int media_changed;
-    int is_read;
+    enum ide_dma_cmd dma_cmd;
     /* SMART */
     uint8_t smart_enabled;
     uint8_t smart_autosave;
@@ -486,6 +502,8 @@ struct IDEBus {
     uint8_t unit;
     uint8_t cmd;
     qemu_irq irq;
+
+    int error_status;
 };
 
 struct IDEDevice {
@@ -505,10 +523,17 @@ struct IDEDeviceInfo {
 #define BM_STATUS_DMAING 0x01
 #define BM_STATUS_ERROR  0x02
 #define BM_STATUS_INT    0x04
+
+/* FIXME These are not status register bits */
 #define BM_STATUS_DMA_RETRY  0x08
 #define BM_STATUS_PIO_RETRY  0x10
 #define BM_STATUS_RETRY_READ  0x20
 #define BM_STATUS_RETRY_FLUSH 0x40
+#define BM_STATUS_RETRY_TRIM 0x80
+
+#define BM_MIGRATION_COMPAT_STATUS_BITS \
+        (BM_STATUS_DMA_RETRY | BM_STATUS_PIO_RETRY | \
+        BM_STATUS_RETRY_READ | BM_STATUS_RETRY_FLUSH)
 
 #define BM_CMD_START     0x01
 #define BM_CMD_READ      0x08
@@ -575,6 +600,9 @@ void ide_transfer_start(IDEState *s, uint8_t *buf, int size,
                         EndTransferFunc *end_transfer_func);
 void ide_transfer_stop(IDEState *s);
 void ide_set_inactive(IDEState *s);
+BlockDriverAIOCB *ide_issue_trim(BlockDriverState *bs,
+        int64_t sector_num, QEMUIOVector *qiov, int nb_sectors,
+        BlockDriverCompletionFunc *cb, void *opaque);
 
 /* hw/ide/atapi.c */
 void ide_atapi_cmd(IDEState *s);
diff --git a/hw/ide/macio.c b/hw/ide/macio.c
index 7107f6b3c2..7daeb31ec3 100644
--- a/hw/ide/macio.c
+++ b/hw/ide/macio.c
@@ -145,12 +145,21 @@ static void pmac_ide_transfer_cb(void *opaque, int ret)
     io->addr += io->len;
     io->len = 0;
 
-    if (s->is_read)
+    switch (s->dma_cmd) {
+    case IDE_DMA_READ:
         m->aiocb = dma_bdrv_read(s->bs, &s->sg, sector_num,
 		                 pmac_ide_transfer_cb, io);
-    else
+        break;
+    case IDE_DMA_WRITE:
         m->aiocb = dma_bdrv_write(s->bs, &s->sg, sector_num,
 		                  pmac_ide_transfer_cb, io);
+        break;
+    case IDE_DMA_TRIM:
+        m->aiocb = dma_bdrv_io(s->bs, &s->sg, sector_num,
+                               ide_issue_trim, pmac_ide_transfer_cb, s, 1);
+        break;
+    }
+
     if (!m->aiocb)
         pmac_ide_transfer_cb(io, -1);
 }
diff --git a/hw/ide/pci.c b/hw/ide/pci.c
index a4726adbea..9f3050a15e 100644
--- a/hw/ide/pci.c
+++ b/hw/ide/pci.c
@@ -169,7 +169,7 @@ static int bmdma_set_inactive(IDEDMA *dma)
     return 0;
 }
 
-static void bmdma_restart_dma(BMDMAState *bm, int is_read)
+static void bmdma_restart_dma(BMDMAState *bm, enum ide_dma_cmd dma_cmd)
 {
     IDEState *s = bmdma_active_if(bm);
 
@@ -177,33 +177,48 @@ static void bmdma_restart_dma(BMDMAState *bm, int is_read)
     s->io_buffer_index = 0;
     s->io_buffer_size = 0;
     s->nsector = bm->nsector;
-    s->is_read = is_read;
+    s->dma_cmd = dma_cmd;
     bm->cur_addr = bm->addr;
     bm->dma_cb = ide_dma_cb;
     bmdma_start_dma(&bm->dma, s, bm->dma_cb);
 }
 
+/* TODO This should be common IDE code */
 static void bmdma_restart_bh(void *opaque)
 {
     BMDMAState *bm = opaque;
+    IDEBus *bus = bm->bus;
     int is_read;
+    int error_status;
 
     qemu_bh_delete(bm->bh);
     bm->bh = NULL;
 
-    is_read = !!(bm->status & BM_STATUS_RETRY_READ);
+    if (bm->unit == (uint8_t) -1) {
+        return;
+    }
+
+    is_read = !!(bus->error_status & BM_STATUS_RETRY_READ);
 
-    if (bm->status & BM_STATUS_DMA_RETRY) {
-        bm->status &= ~(BM_STATUS_DMA_RETRY | BM_STATUS_RETRY_READ);
-        bmdma_restart_dma(bm, is_read);
-    } else if (bm->status & BM_STATUS_PIO_RETRY) {
-        bm->status &= ~(BM_STATUS_PIO_RETRY | BM_STATUS_RETRY_READ);
+    /* The error status must be cleared before resubmitting the request: The
+     * request may fail again, and this case can only be distinguished if the
+     * called function can set a new error status. */
+    error_status = bus->error_status;
+    bus->error_status = 0;
+
+    if (error_status & BM_STATUS_DMA_RETRY) {
+        if (error_status & BM_STATUS_RETRY_TRIM) {
+            bmdma_restart_dma(bm, IDE_DMA_TRIM);
+        } else {
+            bmdma_restart_dma(bm, is_read ? IDE_DMA_READ : IDE_DMA_WRITE);
+        }
+    } else if (error_status & BM_STATUS_PIO_RETRY) {
         if (is_read) {
             ide_sector_read(bmdma_active_if(bm));
         } else {
             ide_sector_write(bmdma_active_if(bm));
         }
-    } else if (bm->status & BM_STATUS_RETRY_FLUSH) {
+    } else if (error_status & BM_STATUS_RETRY_FLUSH) {
         ide_flush_cache(bmdma_active_if(bm));
     }
 }
@@ -351,6 +366,43 @@ static bool ide_bmdma_current_needed(void *opaque)
     return (bm->cur_prd_len != 0);
 }
 
+static bool ide_bmdma_status_needed(void *opaque)
+{
+    BMDMAState *bm = opaque;
+
+    /* Older versions abused some bits in the status register for internal
+     * error state. If any of these bits are set, we must add a subsection to
+     * transfer the real status register */
+    uint8_t abused_bits = BM_MIGRATION_COMPAT_STATUS_BITS;
+
+    return ((bm->status & abused_bits) != 0);
+}
+
+static void ide_bmdma_pre_save(void *opaque)
+{
+    BMDMAState *bm = opaque;
+    uint8_t abused_bits = BM_MIGRATION_COMPAT_STATUS_BITS;
+
+    bm->migration_compat_status =
+        (bm->status & ~abused_bits) | (bm->bus->error_status & abused_bits);
+}
+
+/* This function accesses bm->bus->error_status which is loaded only after
+ * BMDMA itself. This is why the function is called from ide_pci_post_load
+ * instead of being registered with VMState where it would run too early. */
+static int ide_bmdma_post_load(void *opaque, int version_id)
+{
+    BMDMAState *bm = opaque;
+    uint8_t abused_bits = BM_MIGRATION_COMPAT_STATUS_BITS;
+
+    if (bm->status == 0) {
+        bm->status = bm->migration_compat_status & ~abused_bits;
+        bm->bus->error_status |= bm->migration_compat_status & abused_bits;
+    }
+
+    return 0;
+}
+
 static const VMStateDescription vmstate_bmdma_current = {
     .name = "ide bmdma_current",
     .version_id = 1,
@@ -365,15 +417,26 @@ static const VMStateDescription vmstate_bmdma_current = {
     }
 };
 
+const VMStateDescription vmstate_bmdma_status = {
+    .name ="ide bmdma/status",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .minimum_version_id_old = 1,
+    .fields = (VMStateField []) {
+        VMSTATE_UINT8(status, BMDMAState),
+        VMSTATE_END_OF_LIST()
+    }
+};
 
 static const VMStateDescription vmstate_bmdma = {
     .name = "ide bmdma",
     .version_id = 3,
     .minimum_version_id = 0,
     .minimum_version_id_old = 0,
+    .pre_save  = ide_bmdma_pre_save,
     .fields      = (VMStateField []) {
         VMSTATE_UINT8(cmd, BMDMAState),
-        VMSTATE_UINT8(status, BMDMAState),
+        VMSTATE_UINT8(migration_compat_status, BMDMAState),
         VMSTATE_UINT32(addr, BMDMAState),
         VMSTATE_INT64(sector_num, BMDMAState),
         VMSTATE_UINT32(nsector, BMDMAState),
@@ -385,6 +448,9 @@ static const VMStateDescription vmstate_bmdma = {
             .vmsd = &vmstate_bmdma_current,
             .needed = ide_bmdma_current_needed,
         }, {
+            .vmsd = &vmstate_bmdma_status,
+            .needed = ide_bmdma_status_needed,
+        }, {
             /* empty */
         }
     }
@@ -399,7 +465,9 @@ static int ide_pci_post_load(void *opaque, int version_id)
         /* current versions always store 0/1, but older version
            stored bigger values. We only need last bit */
         d->bmdma[i].unit &= 1;
+        ide_bmdma_post_load(&d->bmdma[i], -1);
     }
+
     return 0;
 }
 
diff --git a/hw/ide/pci.h b/hw/ide/pci.h
index cd72cbaeb9..b4f3691a5c 100644
--- a/hw/ide/pci.h
+++ b/hw/ide/pci.h
@@ -22,6 +22,10 @@ typedef struct BMDMAState {
     IORange addr_ioport;
     QEMUBH *bh;
     qemu_irq irq;
+
+    /* Bit 0-2 and 7:   BM status register
+     * Bit 3-6:         bus->error_status */
+    uint8_t migration_compat_status;
 } BMDMAState;
 
 typedef struct PCIIDEState {
diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c
index 3f9dc89c6d..d9b8f24bb5 100644
--- a/hw/ide/qdev.c
+++ b/hw/ide/qdev.c
@@ -125,6 +125,11 @@ static int ide_dev_initfn(IDEDevice *dev, IDEDriveKind kind)
     const char *serial;
     DriveInfo *dinfo;
 
+    if (dev->conf.discard_granularity && dev->conf.discard_granularity != 512) {
+        error_report("discard_granularity must be 512 for ide");
+        return -1;
+    }
+
     serial = dev->serial;
     if (!serial) {
         /* try to fall back to value set with legacy -drive serial=... */