ide: add TRIM support

Add support for TRIM sub function of the data set management command,
and wire it up to the qemu discard infrastructure.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Kevin Wolf <kwolf@redhat.com>

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;
@@ -474,6 +548,9 @@ handle_rw_error:
 
         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;
         }
@@ -519,6 +596,10 @@ handle_rw_error:
         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) {
@@ -818,6 +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)

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
@@ -382,6 +389,7 @@ struct unreported_events {
 enum ide_dma_cmd {
     IDE_DMA_READ,
     IDE_DMA_WRITE,
+    IDE_DMA_TRIM,
 };
 
 #define ide_cmd_is_read(s) \
@@ -521,6 +529,7 @@ struct IDEDeviceInfo {
 #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 | \
@@ -591,6 +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);

hw/ide/macio.c

@@ -154,6 +154,10 @@ static void pmac_ide_transfer_cb(void *opaque, int ret)
         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)

hw/ide/pci.c

@@ -200,8 +200,13 @@ static void bmdma_restart_bh(void *opaque)
     is_read = !!(bus->error_status & BM_STATUS_RETRY_READ);
 
     if (bus->error_status & BM_STATUS_DMA_RETRY) {
+        if (bus->error_status & BM_STATUS_RETRY_TRIM) {
+            bus->error_status &= ~BM_STATUS_RETRY_TRIM;
+            bmdma_restart_dma(bm, IDE_DMA_TRIM);
+        } else {
             bus->error_status &= ~(BM_STATUS_DMA_RETRY | BM_STATUS_RETRY_READ);
             bmdma_restart_dma(bm, is_read ? IDE_DMA_READ : IDE_DMA_WRITE);
+        }
     } else if (bus->error_status & BM_STATUS_PIO_RETRY) {
         bus->error_status &= ~(BM_STATUS_PIO_RETRY | BM_STATUS_RETRY_READ);
         if (is_read) {

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=... */