Skip to contentSkip to conversation

mmc: sdhci: Check for reset prior to DMA address unmap

PATCH In review Pradeep P V K 3 participants 4 messages Mar 3, 2021 – Mar 4, 2021

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 mmc: sdhci: Check for reset prior to DMA address unmap

From: Pradeep P V K <[email protected]>

For data read commands, SDHC may initiate data transfers even before it completely process the command response. In case command itself fails, driver un-maps the memory associated with data transfer but this memory can still be accessed by SDHC for the already initiated data transfer. This scenario can lead to un-mapped memory access error.

To avoid this scenario, reset SDHC (when command fails) prior to un-mapping memory. Resetting SDHC ensures that all in-flight data transfers are either aborted or completed. So we don't run into this scenario.

Swap the reset, un-map steps sequence in sdhci_request_done().

Changes since V1:
- Added an empty line and fixed the comment style.
- Retained the Acked-by signoff.

---
 drivers/mmc/host/sdhci.c | 60 +++++++++++++++++++++++++-----------------------
 1 file changed, 31 insertions(+), 29 deletions(-)
drivers/mmc/host/sdhci.c +31 -29
index 646823d..130fd2d 100644
@@ -2998,6 +2998,37 @@ static bool sdhci_request_done(struct sdhci_host *host)
 	}
 
 	/*
+	 * The controller needs a reset of internal state machines
+	 * upon error conditions.
+	 */
+	if (sdhci_needs_reset(host, mrq)) {
+		/*
+		 * Do not finish until command and data lines are available for
+		 * reset. Note there can only be one other mrq, so it cannot
+		 * also be in mrqs_done, otherwise host->cmd and host->data_cmd
+		 * would both be null.
+		 */
+		if (host->cmd || host->data_cmd) {
+			spin_unlock_irqrestore(&host->lock, flags);
+			return true;
+		}
+
+		/* Some controllers need this kick or reset won't work here */
+		if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
+			/* This is to force an update */
+			host->ops->set_clock(host, host->clock);
+
+		/*
+		 * Spec says we should do both at the same time, but Ricoh
+		 * controllers do not like that.
+		 */
+		sdhci_do_reset(host, SDHCI_RESET_CMD);
+		sdhci_do_reset(host, SDHCI_RESET_DATA);
+
+		host->pending_reset = false;
+	}
+
+	/*
 	 * Always unmap the data buffers if they were mapped by
 	 * sdhci_prepare_data() whenever we finish with a request.
 	 * This avoids leaking DMA mappings on error.
@@ -3060,35 +3091,6 @@ static bool sdhci_request_done(struct sdhci_host *host)
 		}
 	}
 
-	/*
-	 * The controller needs a reset of internal state machines
-	 * upon error conditions.
-	 */
-	if (sdhci_needs_reset(host, mrq)) {
-		/*
-		 * Do not finish until command and data lines are available for
-		 * reset. Note there can only be one other mrq, so it cannot
-		 * also be in mrqs_done, otherwise host->cmd and host->data_cmd
-		 * would both be null.
-		 */
-		if (host->cmd || host->data_cmd) {
-			spin_unlock_irqrestore(&host->lock, flags);
-			return true;
-		}
-
-		/* Some controllers need this kick or reset won't work here */
-		if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
-			/* This is to force an update */
-			host->ops->set_clock(host, host->clock);
-
-		/* Spec says we should do both at the same time, but Ricoh
-		   controllers do not like that. */
-		sdhci_do_reset(host, SDHCI_RESET_CMD);
-		sdhci_do_reset(host, SDHCI_RESET_DATA);
-
-		host->pending_reset = false;
-	}
-
 	host->mrqs_done[i] = NULL;
 
 	spin_unlock_irqrestore(&host->lock, flags);
Signature · 1 line
2.7.4
#1 lore raw
Quoted 16 lines · Pradeep P V K

From: Pradeep P V K <[email protected]>

For data read commands, SDHC may initiate data transfers even before it completely process the command response. In case command itself fails, driver un-maps the memory associated with data transfer but this memory can still be accessed by SDHC for the already initiated data transfer. This scenario can lead to un-mapped memory access error.

To avoid this scenario, reset SDHC (when command fails) prior to un-mapping memory. Resetting SDHC ensures that all in-flight data transfers are either aborted or completed. So we don't run into this scenario.

Swap the reset, un-map steps sequence in sdhci_request_done().

Quoted 7 lines

Changes since V1:
- Added an empty line and fixed the comment style.
- Retained the Acked-by signoff.

Seems like it might be a good idea to tag this for stable? I did that, but awaiting for your confirmation.

So, applied for next, thanks!

Kind regards
Uffe

Quoted 85 lines
---
 drivers/mmc/host/sdhci.c | 60 +++++++++++++++++++++++++-----------------------
 1 file changed, 31 insertions(+), 29 deletions(-)
drivers/mmc/host/sdhci.c +31 -30
index 646823d..130fd2d 100644
@@ -2998,6 +2998,37 @@ static bool sdhci_request_done(struct sdhci_host *host)
        }
 
        /*
+        * The controller needs a reset of internal state machines
+        * upon error conditions.
+        */
+       if (sdhci_needs_reset(host, mrq)) {
+               /*
+                * Do not finish until command and data lines are available for
+                * reset. Note there can only be one other mrq, so it cannot
+                * also be in mrqs_done, otherwise host->cmd and host->data_cmd
+                * would both be null.
+                */
+               if (host->cmd || host->data_cmd) {
+                       spin_unlock_irqrestore(&host->lock, flags);
+                       return true;
+               }
+
+               /* Some controllers need this kick or reset won't work here */
+               if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
+                       /* This is to force an update */
+                       host->ops->set_clock(host, host->clock);
+
+               /*
+                * Spec says we should do both at the same time, but Ricoh
+                * controllers do not like that.
+                */
+               sdhci_do_reset(host, SDHCI_RESET_CMD);
+               sdhci_do_reset(host, SDHCI_RESET_DATA);
+
+               host->pending_reset = false;
+       }
+
+       /*
         * Always unmap the data buffers if they were mapped by
         * sdhci_prepare_data() whenever we finish with a request.
         * This avoids leaking DMA mappings on error.
@@ -3060,35 +3091,6 @@ static bool sdhci_request_done(struct sdhci_host *host)
                }
        }
 
-       /*
-        * The controller needs a reset of internal state machines
-        * upon error conditions.
-        */
-       if (sdhci_needs_reset(host, mrq)) {
-               /*
-                * Do not finish until command and data lines are available for
-                * reset. Note there can only be one other mrq, so it cannot
-                * also be in mrqs_done, otherwise host->cmd and host->data_cmd
-                * would both be null.
-                */
-               if (host->cmd || host->data_cmd) {
-                       spin_unlock_irqrestore(&host->lock, flags);
-                       return true;
-               }
-
-               /* Some controllers need this kick or reset won't work here */
-               if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
-                       /* This is to force an update */
-                       host->ops->set_clock(host, host->clock);
-
-               /* Spec says we should do both at the same time, but Ricoh
-                  controllers do not like that. */
-               sdhci_do_reset(host, SDHCI_RESET_CMD);
-               sdhci_do_reset(host, SDHCI_RESET_DATA);
-
-               host->pending_reset = false;
-       }
-
        host->mrqs_done[i] = NULL;
 
        spin_unlock_irqrestore(&host->lock, flags);
--

2.7.4

#2 lore raw
Quoted 17 lines · Ulf Hansson
Quoted 16 lines · Pradeep P V K

From: Pradeep P V K <[email protected]>

For data read commands, SDHC may initiate data transfers even before it completely process the command response. In case command itself fails, driver un-maps the memory associated with data transfer but this memory can still be accessed by SDHC for the already initiated data transfer. This scenario can lead to un-mapped memory access error.

To avoid this scenario, reset SDHC (when command fails) prior to un-mapping memory. Resetting SDHC ensures that all in-flight data transfers are either aborted or completed. So we don't run into this scenario.

Swap the reset, un-map steps sequence in sdhci_request_done().

Quoted 7 lines
Quoted 7 lines

Changes since V1:
- Added an empty line and fixed the comment style.
- Retained the Acked-by signoff.

Hi Uffe,

Quoted 4 lines

Seems like it might be a good idea to tag this for stable? I did that, but awaiting for your confirmation.

Yes, this fix is applicable for all stable starting from 4.9 (n/a for 4.4). Kindly go ahead.

Quoted 5 lines

So, applied for next, thanks!

Kind regards
Uffe

Thanks and Regards,
Pradeep

Quoted 99 lines
Quoted 99 lines
---
 drivers/mmc/host/sdhci.c | 60 
+++++++++++++++++++++++++-----------------------
 1 file changed, 31 insertions(+), 29 deletions(-)
drivers/mmc/host/sdhci.c +6 -0
index 646823d..130fd2d 100644
@@ -2998,6 +2998,37 @@ static bool sdhci_request_done(struct 
 sdhci_host *host)
        }
 
        /*
+        * The controller needs a reset of internal state machines
+        * upon error conditions.
+        */
+       if (sdhci_needs_reset(host, mrq)) {
+               /*
+                * Do not finish until command and data lines are 

available for
+ * reset. Note there can only be one other mrq, so it
cannot
+ * also be in mrqs_done, otherwise host->cmd and
host->data_cmd

drivers/mmc/host/sdhci.c +8 -0
+                * would both be null.
+                */
+               if (host->cmd || host->data_cmd) {
+                       spin_unlock_irqrestore(&host->lock, flags);
+                       return true;
+               }
+
+               /* Some controllers need this kick or reset won't work 

here */

drivers/mmc/host/sdhci.c +6 -0
+               if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
+                       /* This is to force an update */
+                       host->ops->set_clock(host, host->clock);
+
+               /*
+                * Spec says we should do both at the same time, but 

Ricoh

drivers/mmc/host/sdhci.c +9 -7
+                * controllers do not like that.
+                */
+               sdhci_do_reset(host, SDHCI_RESET_CMD);
+               sdhci_do_reset(host, SDHCI_RESET_DATA);
+
+               host->pending_reset = false;
+       }
+
+       /*
         * Always unmap the data buffers if they were mapped by
         * sdhci_prepare_data() whenever we finish with a request.
         * This avoids leaking DMA mappings on error.
@@ -3060,35 +3091,6 @@ static bool sdhci_request_done(struct 
 sdhci_host *host)
                }
        }
 
-       /*
-        * The controller needs a reset of internal state machines
-        * upon error conditions.
-        */
-       if (sdhci_needs_reset(host, mrq)) {
-               /*
-                * Do not finish until command and data lines are 

available for
- * reset. Note there can only be one other mrq, so it
cannot
- * also be in mrqs_done, otherwise host->cmd and
host->data_cmd

drivers/mmc/host/sdhci.c +0 -8
-                * would both be null.
-                */
-               if (host->cmd || host->data_cmd) {
-                       spin_unlock_irqrestore(&host->lock, flags);
-                       return true;
-               }
-
-               /* Some controllers need this kick or reset won't work 

here */

drivers/mmc/host/sdhci.c +0 -5
-               if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
-                       /* This is to force an update */
-                       host->ops->set_clock(host, host->clock);
-
-               /* Spec says we should do both at the same time, but 

Ricoh

drivers/mmc/host/sdhci.c +0 -8
-                  controllers do not like that. */
-               sdhci_do_reset(host, SDHCI_RESET_CMD);
-               sdhci_do_reset(host, SDHCI_RESET_DATA);
-
-               host->pending_reset = false;
-       }
-
        host->mrqs_done[i] = NULL;
 
        spin_unlock_irqrestore(&host->lock, flags);
--

2.7.4

#3 lore raw
Quoted 39 lines · On Thu, 4 Mar 2021 at 16:16, <[email protected]> wrote:
Quoted 20 lines · Ulf Hansson
Quoted 18 lines · Pradeep P V K

From: Pradeep P V K <[email protected]>

For data read commands, SDHC may initiate data transfers even before it
completely process the command response. In case command itself fails, driver un-maps the memory associated with data transfer but this memory
can still be accessed by SDHC for the already initiated data transfer. This scenario can lead to un-mapped memory access error.

To avoid this scenario, reset SDHC (when command fails) prior to un-mapping memory. Resetting SDHC ensures that all in-flight data transfers are either aborted or completed. So we don't run into this scenario.

Swap the reset, un-map steps sequence in sdhci_request_done().

Quoted 7 lines
Quoted 7 lines

Changes since V1:
- Added an empty line and fixed the comment style.
- Retained the Acked-by signoff.

Hi Uffe,

Quoted 4 lines

Seems like it might be a good idea to tag this for stable? I did that, but awaiting for your confirmation.

Yes, this fix is applicable for all stable starting from 4.9 (n/a for 4.4).
Kindly go ahead.

Quoted 7 lines
Quoted 5 lines

So, applied for next, thanks!

Kind regards
Uffe

Thanks and Regards,
Pradeep

Thanks for confirming, I have updated the stable tag.

Kind regards
Uffe

Quoted 100 lines
Quoted 100 lines
Quoted 100 lines

---
 drivers/mmc/host/sdhci.c | 60
+++++++++++++++++++++++++-----------------------
 1 file changed, 31 insertions(+), 29 deletions(-)
drivers/mmc/host/sdhci.c +6 -0
index 646823d..130fd2d 100644
@@ -2998,6 +2998,37 @@ static bool sdhci_request_done(struct
 sdhci_host *host)
        }
 
        /*
+        * The controller needs a reset of internal state machines
+        * upon error conditions.
+        */
+       if (sdhci_needs_reset(host, mrq)) {
+               /*
+                * Do not finish until command and data lines are

available for
+ * reset. Note there can only be one other mrq, so it
cannot
+ * also be in mrqs_done, otherwise host->cmd and
host->data_cmd

drivers/mmc/host/sdhci.c +8 -0
+                * would both be null.
+                */
+               if (host->cmd || host->data_cmd) {
+                       spin_unlock_irqrestore(&host->lock, flags);
+                       return true;
+               }
+
+               /* Some controllers need this kick or reset won't work

here */

drivers/mmc/host/sdhci.c +6 -0
+               if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
+                       /* This is to force an update */
+                       host->ops->set_clock(host, host->clock);
+
+               /*
+                * Spec says we should do both at the same time, but

Ricoh

drivers/mmc/host/sdhci.c +9 -7
+                * controllers do not like that.
+                */
+               sdhci_do_reset(host, SDHCI_RESET_CMD);
+               sdhci_do_reset(host, SDHCI_RESET_DATA);
+
+               host->pending_reset = false;
+       }
+
+       /*
         * Always unmap the data buffers if they were mapped by
         * sdhci_prepare_data() whenever we finish with a request.
         * This avoids leaking DMA mappings on error.
@@ -3060,35 +3091,6 @@ static bool sdhci_request_done(struct
 sdhci_host *host)
                }
        }
 
-       /*
-        * The controller needs a reset of internal state machines
-        * upon error conditions.
-        */
-       if (sdhci_needs_reset(host, mrq)) {
-               /*
-                * Do not finish until command and data lines are

available for
- * reset. Note there can only be one other mrq, so it
cannot
- * also be in mrqs_done, otherwise host->cmd and
host->data_cmd

drivers/mmc/host/sdhci.c +0 -8
-                * would both be null.
-                */
-               if (host->cmd || host->data_cmd) {
-                       spin_unlock_irqrestore(&host->lock, flags);
-                       return true;
-               }
-
-               /* Some controllers need this kick or reset won't work

here */

drivers/mmc/host/sdhci.c +0 -5
-               if (host->quirks & SDHCI_QUIRK_CLOCK_BEFORE_RESET)
-                       /* This is to force an update */
-                       host->ops->set_clock(host, host->clock);
-
-               /* Spec says we should do both at the same time, but

Ricoh

drivers/mmc/host/sdhci.c +0 -8
-                  controllers do not like that. */
-               sdhci_do_reset(host, SDHCI_RESET_CMD);
-               sdhci_do_reset(host, SDHCI_RESET_DATA);
-
-               host->pending_reset = false;
-       }
-
        host->mrqs_done[i] = NULL;
 
        spin_unlock_irqrestore(&host->lock, flags);
--

2.7.4