Skip to contentSkip to conversation

SD and SDHCI Fixes

PATCH Superseded [email protected] 2 participants 8 messages May 24, 2013

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover SD and SDHCI Fixes

From: Peter Crosthwaite <[email protected]>

Fixes found in SD and SDHCI found doing some corner case testing.

Changed from v1:
Corrected ACMD41 patch.
Moved ACMD41 patch to back of series.

Peter Crosthwaite (5):
sd/sdhci.c: Only reset data_count on new commands sd/sdhci: Fix Buffer Write Ready interrupt sd/sdhci.c: Fix bdata_read DPRINT message sd/sdhci:ADMA: fix interrupt
sd/sd.c: Fix "inquiry" ACMD41

 hw/sd/sd.c    |  3 ++-
 hw/sd/sdhci.c | 28 ++++++++++++++--------------
 2 files changed, 16 insertions(+), 15 deletions(-)
Signature · 1 line
1.8.3.rc1.44.gb387c77.dirty
#1 lore raw

PATCH v2 1/5 sd/sdhci.c: Only reset data_count on new commands

From: Peter Crosthwaite <[email protected]>

The data_count variable was being reset on every transfer, including DMA transfer resumptions. This is incorrect, it should only be set on a new command.

Manifests as a bug when using ADMA and there is a timer delay between ADMA frames where the fifo is left in a non empty state.

---

 hw/sd/sdhci.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
hw/sd/sdhci.c +1 -1
index 91dc9b0..0a84540 100644
@@ -260,6 +260,7 @@ static void sdhci_send_command(SDHCIState *s)
     sdhci_update_irq(s);
 
     if (s->blksize && (s->cmdreg & SDHC_CMD_DATA_PRESENT)) {
+        s->data_count = 0;
         sdhci_do_data_transfer(s);
     }
 }
@@ -773,7 +774,6 @@ static void sdhci_do_adma(SDHCIState *s)
 static void sdhci_data_transfer(SDHCIState *s)
 {
     SDHCIClass *k = SDHCI_GET_CLASS(s);
-    s->data_count = 0;
 
     if (s->trnmod & SDHC_TRNS_DMA) {
         switch (SDHC_DMA_TYPE(s->hostctl)) {
Signature · 1 line
1.8.3.rc1.44.gb387c77.dirty
#2 lore raw

PATCH v2 2/5 sd/sdhci: Fix Buffer Write Ready interrupt

In reply to [email protected]

From: Peter Crosthwaite <[email protected]>

This interrupt is not risen after the last block is written to sd. It is mutually exclusive with the end of transfer conditions. Fix.

---

 hw/sd/sdhci.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)
hw/sd/sdhci.c +2 -3
index 0a84540..ea510b5 100644
@@ -405,15 +405,14 @@ static void sdhci_write_block_to_card(SDHCIState *s)
 
     /* Next data can be written through BUFFER DATORT register */
     s->prnsts |= SDHC_SPACE_AVAILABLE;
-    if (s->norintstsen & SDHC_NISEN_WBUFRDY) {
-        s->norintsts |= SDHC_NIS_WBUFRDY;
-    }
 
     /* Finish transfer if that was the last block of data */
     if ((s->trnmod & SDHC_TRNS_MULTI) == 0 ||
             ((s->trnmod & SDHC_TRNS_MULTI) &&
             (s->trnmod & SDHC_TRNS_BLK_CNT_EN) && (s->blkcnt == 0))) {
         SDHCI_GET_CLASS(s)->end_data_transfer(s);
+    } else if (s->norintstsen & SDHC_NISEN_WBUFRDY) {
+        s->norintsts |= SDHC_NIS_WBUFRDY;
     }
 
     /* Generate Block Gap Event if requested and if not the last block */
Signature · 1 line
1.8.3.rc1.44.gb387c77.dirty
#3 lore raw

PATCH v2 3/5 sd/sdhci.c: Fix bdata_read DPRINT message

In reply to [email protected]

From: Peter Crosthwaite <[email protected]>

This message was printing out the data in decimal only, which is not very friendly to the debugging developer. Add hex variant in parenthesis to make it consistent with other similar messages in this module.

---

 hw/sd/sdhci.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
hw/sd/sdhci.c +2 -1
index ea510b5..15345dc 100644
@@ -880,7 +880,8 @@ static uint32_t sdhci_read(SDHCIState *s, unsigned int offset, unsigned size)
     case  SDHC_BDATA:
         if (sdhci_buff_access_is_sequential(s, offset - SDHC_BDATA)) {
             ret = SDHCI_GET_CLASS(s)->bdata_read(s, size);
-            DPRINT_L2("read %ub: addr[0x%04x] -> %u\n", size, offset, ret);
+            DPRINT_L2("read %ub: addr[0x%04x] -> %u(0x%x)\n", size, offset,
+                      ret, ret);
             return ret;
         }
         break;
Signature · 1 line
1.8.3.rc1.44.gb387c77.dirty
#4 lore raw

PATCH v2 4/5 sd/sdhci:ADMA: fix interrupt

In reply to [email protected]

From: Peter Crosthwaite <[email protected]>

The end of transfer check was occurring and potentially returning before the interrupt flag was checked. This means the interrupt will be missed if it occurs on the last packet. Fix by checking for the interrupt before checking for the end of transfer.

---

 hw/sd/sdhci.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)
hw/sd/sdhci.c +9 -9
index 15345dc..e64899c 100644
@@ -730,6 +730,15 @@ static void sdhci_do_adma(SDHCIState *s)
             break;
         }
 
+        if (dscr.attr & SDHC_ADMA_ATTR_INT) {
+            DPRINT_L1("ADMA interrupt: admasysaddr=0x%lx\n", s->admasysaddr);
+            if (s->norintstsen & SDHC_NISEN_DMA) {
+                s->norintsts |= SDHC_NIS_DMA;
+            }
+
+            sdhci_update_irq(s);
+        }
+
         /* ADMA transfer terminates if blkcnt == 0 or by END attribute */
         if (((s->trnmod & SDHC_TRNS_BLK_CNT_EN) &&
                     (s->blkcnt == 0)) || (dscr.attr & SDHC_ADMA_ATTR_END)) {
@@ -752,15 +761,6 @@ static void sdhci_do_adma(SDHCIState *s)
             return;
         }
 
-        if (dscr.attr & SDHC_ADMA_ATTR_INT) {
-            DPRINT_L1("ADMA interrupt: admasysaddr=0x%lx\n", s->admasysaddr);
-            if (s->norintstsen & SDHC_NISEN_DMA) {
-                s->norintsts |= SDHC_NIS_DMA;
-            }
-
-            sdhci_update_irq(s);
-            return;
-        }
     }
 
     /* we have unfinished business - reschedule to continue ADMA */
Signature · 1 line
1.8.3.rc1.44.gb387c77.dirty
#5 lore raw

PATCH v2 5/5 sd/sd.c: Fix "inquiry" ACMD41

In reply to [email protected]

From: Peter Crosthwaite <[email protected]>

QEMU models two (of the three) ACMD41 has two modes, "inquiry" and "first". The selection logic for which of the two is incorrect - it compares != 0 for the entire argument value rather than only bits 23:0 as per the spec. Fix.

---
Changed since v1:
Total rewrite

 hw/sd/sd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
hw/sd/sd.c +2 -1
index 2e0ef3e..4d26578 100644
@@ -1278,8 +1278,9 @@ static sd_rsp_type_t sd_app_command(SDState *sd,
         switch (sd->state) {
         case sd_idle_state:
             /* We accept any voltage.  10000 V is nothing.  */
-            if (req.arg)
+            if (req.arg & 0x00ffffff) {
                 sd->state = sd_ready_state;
+            }
 
             return sd_r3;
 
Signature · 1 line
1.8.3.rc1.44.gb387c77.dirty
#6 lore raw

Re: [Qemu-devel] [PATCH arm-devs v2 5/5] sd/sd.c: Fix "inquiry" ACMD41

Quoted 20 lines · On 24 May 2013 06:18, <[email protected]> wrote:

From: Peter Crosthwaite <[email protected]>

QEMU models two (of the three) ACMD41 has two modes, "inquiry" and "first". The selection logic for which of the two is incorrect - it compares != 0 for the entire argument value rather than only bits 23:0 as per the spec. Fix.

---
Changed since v1:
Total rewrite

 hw/sd/sd.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
hw/sd/sd.c +0 -0
index 2e0ef3e..4d26578 100644

hw/sd/sd.c static sd_rsp_type_t sd_app_command(SDState *sd, · line 1285

Quoted 9 lines
hw/sd/sd.c +2 -1
@@ -1278,8 +1278,9 @@ static sd_rsp_type_t sd_app_command(SDState *sd,
         switch (sd->state) {
         case sd_idle_state:
             /* We accept any voltage.  10000 V is nothing.  */
-            if (req.arg)
+            if (req.arg & 0x00ffffff) {
                 sd->state = sd_ready_state;
+            }
 
             return sd_r3;

Could we have a comment here please?

thanks
-- PMM