Skip to contentSkip to conversation

mmc_spi: Fix cd-gpios for SPI SD card insertion/removal detection

PATCH In review cuu 2 participants 6 messages Aug 24, 2025 – Oct 23, 2025 1 message not in the archive MMC, SD and SDIO

lore
Share & tools
lore

cuu

author
#0 lore raw

PATCH 1/1 mmc_spi: Fix cd-gpios for SPI SD card insertion/removal detection

Previously, SD cards using pure SPI protocol with mmc_spi could not reliably detect card removal via cd-gpios. This patch fixes the issue by extending the card alive check in _mmc_detect_card_removed() to include cd-gpios detection.

Additionally, mmc_rescan() now includes a status probing function specific to SPI SD cards to ensure the card state is correctly updated on insertion and removal events.

---
 drivers/mmc/core/core.c | 15 ++++++++++++---
 drivers/mmc/core/core.h |  1 +
 drivers/mmc/core/mmc.c  |  8 ++++++++
 3 files changed, 21 insertions(+), 3 deletions(-)
drivers/mmc/core/core.c +12 -3
index 874c6fe92855..e028ec8fbf46 100644
@@ -2123,6 +2123,11 @@ int _mmc_detect_card_removed(struct mmc_host *host)
 
 	ret = host->bus_ops->alive(host);
 
+        if(mmc_host_is_spi(host)) {
+        	if(host->ops->get_cd){
+       			ret = !host->ops->get_cd(host);
+                }
+        }
 	/*
 	 * Card detect status and alive check may be out of sync if card is
 	 * removed slowly, when card detect switch changes while card/slot
@@ -2238,9 +2243,13 @@ void mmc_rescan(struct work_struct *work)
 	}
 
 	/* Verify a registered card to be functional, else remove it. */
-	if (host->bus_ops)
-		host->bus_ops->detect(host);
-
+	if (host->bus_ops){
+		if (mmc_host_is_spi(host)){
+			mmc_detect_mmc(host);
+		}else{
+			host->bus_ops->detect(host);
+		}
+	}
 	host->detect_change = 0;
 
 	/* if there still is a card present, stop here */
drivers/mmc/core/core.h +1 -0
index 622085cd766f..5b48c50b5bb5 100644
@@ -78,6 +78,7 @@ void _mmc_detect_change(struct mmc_host *host, unsigned long delay,
 int _mmc_detect_card_removed(struct mmc_host *host);
 int mmc_detect_card_removed(struct mmc_host *host);
 
+int mmc_detect_mmc(struct mmc_host *host);
 int mmc_attach_mmc(struct mmc_host *host);
 int mmc_attach_sd(struct mmc_host *host);
 int mmc_attach_sdio(struct mmc_host *host);
drivers/mmc/core/mmc.c +8 -0
index 5be9b42d5057..e73de34e072b 100644
@@ -2316,6 +2316,14 @@ static const struct mmc_bus_ops mmc_ops = {
 	.flush_cache = _mmc_flush_cache,
 };
 
+/*
+ * Export function for mmc detection 
+ */
+int mmc_detect_mmc(struct mmc_host *host){
+	mmc_detect(host);
+	return 0;
+}
+
 /*
  * Starting point for MMC card init.
  */
Signature · 1 line
2.34.1
#1 lore raw
Quoted 9 lines · cuu

Previously, SD cards using pure SPI protocol with mmc_spi could not reliably detect card removal via cd-gpios. This patch fixes the issue by extending the card alive check in _mmc_detect_card_removed() to include cd-gpios detection.

Additionally, mmc_rescan() now includes a status probing function specific to SPI SD cards to ensure the card state is correctly updated on insertion and removal events.

Sorry, I don't get this.

What is so special with SPI cards here that makes the regular card detection/removal procedures not to work?

Is there something wrong with mmc_sd_alive()?

Quoted 2 lines

Please use a proper name for your sob line. I assume "cuu" is an alias, no?

Kind regards
Uffe

Quoted 73 lines
---
 drivers/mmc/core/core.c | 15 ++++++++++++---
 drivers/mmc/core/core.h |  1 +
 drivers/mmc/core/mmc.c  |  8 ++++++++
 3 files changed, 21 insertions(+), 3 deletions(-)
drivers/mmc/core/core.c +12 -3
index 874c6fe92855..e028ec8fbf46 100644
@@ -2123,6 +2123,11 @@ int _mmc_detect_card_removed(struct mmc_host *host)
 
        ret = host->bus_ops->alive(host);
 
+        if(mmc_host_is_spi(host)) {
+               if(host->ops->get_cd){
+                               ret = !host->ops->get_cd(host);
+                }
+        }
        /*
         * Card detect status and alive check may be out of sync if card is
         * removed slowly, when card detect switch changes while card/slot
@@ -2238,9 +2243,13 @@ void mmc_rescan(struct work_struct *work)
        }
 
        /* Verify a registered card to be functional, else remove it. */
-       if (host->bus_ops)
-               host->bus_ops->detect(host);
-
+       if (host->bus_ops){
+               if (mmc_host_is_spi(host)){
+                       mmc_detect_mmc(host);
+               }else{
+                       host->bus_ops->detect(host);
+               }
+       }
        host->detect_change = 0;
 
        /* if there still is a card present, stop here */
drivers/mmc/core/core.h +1 -0
index 622085cd766f..5b48c50b5bb5 100644
@@ -78,6 +78,7 @@ void _mmc_detect_change(struct mmc_host *host, unsigned long delay,
 int _mmc_detect_card_removed(struct mmc_host *host);
 int mmc_detect_card_removed(struct mmc_host *host);
 
+int mmc_detect_mmc(struct mmc_host *host);
 int mmc_attach_mmc(struct mmc_host *host);
 int mmc_attach_sd(struct mmc_host *host);
 int mmc_attach_sdio(struct mmc_host *host);
drivers/mmc/core/mmc.c +8 -1
index 5be9b42d5057..e73de34e072b 100644
@@ -2316,6 +2316,14 @@ static const struct mmc_bus_ops mmc_ops = {
        .flush_cache = _mmc_flush_cache,
 };
 
+/*
+ * Export function for mmc detection
+ */
+int mmc_detect_mmc(struct mmc_host *host){
+       mmc_detect(host);
+       return 0;
+}
+
 /*
  * Starting point for MMC card init.
  */
--

2.34.1

A message that is not in the archive (sent off-list or not yet received): nyiodgyzyb6f2kkjlvjdlfflfkdccr2lgnij2uxkdm4mrpznf5@srcxlmdh47da

author
#3 lore raw
Quoted 161 lines · cuu
Quoted 85 lines · Ulf Hansson
Quoted 9 lines · cuu

Previously, SD cards using pure SPI protocol with mmc_spi could not reliably detect card removal via cd-gpios. This patch fixes the issue by extending the card alive check in _mmc_detect_card_removed() to include cd-gpios detection.

Additionally, mmc_rescan() now includes a status probing function specific to SPI SD cards to ensure the card state is correctly updated on insertion and removal events.

Sorry, I don't get this.

What is so special with SPI cards here that makes the regular card detection/removal procedures not to work?

Is there something wrong with mmc_sd_alive()?

Quoted 2 lines

Please use a proper name for your sob line. I assume "cuu" is an alias, no?

Kind regards
Uffe

Quoted 59 lines
---
 drivers/mmc/core/core.c | 15 ++++++++++++---
 drivers/mmc/core/core.h |  1 +
 drivers/mmc/core/mmc.c  |  8 ++++++++
 3 files changed, 21 insertions(+), 3 deletions(-)
drivers/mmc/core/core.c +12 -3
index 874c6fe92855..e028ec8fbf46 100644
@@ -2123,6 +2123,11 @@ int _mmc_detect_card_removed(struct mmc_host *host)
 
        ret = host->bus_ops->alive(host);
 
+        if(mmc_host_is_spi(host)) {
+               if(host->ops->get_cd){
+                               ret = !host->ops->get_cd(host);
+                }
+        }
        /*
         * Card detect status and alive check may be out of sync if card is
         * removed slowly, when card detect switch changes while card/slot
@@ -2238,9 +2243,13 @@ void mmc_rescan(struct work_struct *work)
        }
 
        /* Verify a registered card to be functional, else remove it. */
-       if (host->bus_ops)
-               host->bus_ops->detect(host);
-
+       if (host->bus_ops){
+               if (mmc_host_is_spi(host)){
+                       mmc_detect_mmc(host);
+               }else{
+                       host->bus_ops->detect(host);
+               }
+       }
        host->detect_change = 0;
 
        /* if there still is a card present, stop here */
drivers/mmc/core/core.h +1 -0
index 622085cd766f..5b48c50b5bb5 100644
@@ -78,6 +78,7 @@ void _mmc_detect_change(struct mmc_host *host, unsigned long delay,
 int _mmc_detect_card_removed(struct mmc_host *host);
 int mmc_detect_card_removed(struct mmc_host *host);
 
+int mmc_detect_mmc(struct mmc_host *host);
 int mmc_attach_mmc(struct mmc_host *host);
 int mmc_attach_sd(struct mmc_host *host);
 int mmc_attach_sdio(struct mmc_host *host);
drivers/mmc/core/mmc.c +0 -0
index 5be9b42d5057..e73de34e072b 100644
@@ -2316,6 +2316,14 @@ static const struct mmc_bus_ops mmc_ops = {
        .flush_cache = _mmc_flush_cache,
 };
 
Quoted 14 lines
Quoted 12 lines
drivers/mmc/core/mmc.c +8 -1
+/*
+ * Export function for mmc detection
+ */
+int mmc_detect_mmc(struct mmc_host *host){
+       mmc_detect(host);
+       return 0;
+}
+
 /*
  * Starting point for MMC card init.
  */
--
Quoted 2 lines

2.34.1

Hi Ulf

I worked On a PicoCalc board I added an extension board for the Raspberry Pi Zero 2W, while still wanting the Zero 2W to access the PicoCalc’s own SPI-connected SD card.

Here is the DTS I wrote:

```
/dts-v1/;
/plugin/;

/ {
    compatible = "brcm,bcm2835";

    fragment@0 {
        target = <&spidev0>;
        __overlay__ {
            status = "disabled";
        };
    };

    fragment@1 {
        target = <&spi1>;
        __overlay__ {
            status = "okay";

            #address-cells = <1>;
            #size-cells = <0>;

            sdslot@0 {
                compatible = "mmc-spi-slot";
                reg = <0>;                        // SPI CS0
                voltage-ranges = <3300 3300>;
                spi-max-frequency = <8000000>;
                gpios = <&gpio 6 1>;
                cd-gpios = <&gpio 6 1>;
            };
        };
    };
};
```

With this setup, card insertion is detected, but removal is not. From what I can tell, mmc_sd_alive() itself is fine, but it seems does not check the state of cd-gpios.

For removable SPI SD cards that provide cd-gpios, I believe relying on the cd-gpios state may be a more direct and accurate way to determine card presence.

Therefore, I added a check in _mmc_detect_card_removed() to handle the case of SPI-protocol cards with cd-gpios, so that removal detection can work properly.

I am not deeply familiar with the Linux kernel, so if I am wrong please correct me. Thanks a lot.

Ps:

You are right and I am sorry, my sob line was wrong the correct version is

Quoted 9 lines
```
Signed-off-by: dexter kidd <[email protected]>
```

I don't know how to fix the sob line, re-send a patch to you again?

Kind regards

Resending my previous reply with the full CC list this time. Apologies for the noise.

#4 lore raw
Quoted 131 lines · cuu
Quoted 129 lines · cuu
Quoted 85 lines · Ulf Hansson
Quoted 9 lines · cuu

Previously, SD cards using pure SPI protocol with mmc_spi could not reliably detect card removal via cd-gpios. This patch fixes the issue by extending the card alive check in _mmc_detect_card_removed() to include cd-gpios detection.

Additionally, mmc_rescan() now includes a status probing function specific to SPI SD cards to ensure the card state is correctly updated on insertion and removal events.

Sorry, I don't get this.

What is so special with SPI cards here that makes the regular card detection/removal procedures not to work?

Is there something wrong with mmc_sd_alive()?

Quoted 2 lines

Please use a proper name for your sob line. I assume "cuu" is an alias, no?

Kind regards
Uffe

Quoted 59 lines
---
 drivers/mmc/core/core.c | 15 ++++++++++++---
 drivers/mmc/core/core.h |  1 +
 drivers/mmc/core/mmc.c  |  8 ++++++++
 3 files changed, 21 insertions(+), 3 deletions(-)
drivers/mmc/core/core.c +12 -3
index 874c6fe92855..e028ec8fbf46 100644
@@ -2123,6 +2123,11 @@ int _mmc_detect_card_removed(struct mmc_host *host)
 
        ret = host->bus_ops->alive(host);
 
+        if(mmc_host_is_spi(host)) {
+               if(host->ops->get_cd){
+                               ret = !host->ops->get_cd(host);
+                }
+        }
        /*
         * Card detect status and alive check may be out of sync if card is
         * removed slowly, when card detect switch changes while card/slot
@@ -2238,9 +2243,13 @@ void mmc_rescan(struct work_struct *work)
        }
 
        /* Verify a registered card to be functional, else remove it. */
-       if (host->bus_ops)
-               host->bus_ops->detect(host);
-
+       if (host->bus_ops){
+               if (mmc_host_is_spi(host)){
+                       mmc_detect_mmc(host);
+               }else{
+                       host->bus_ops->detect(host);
+               }
+       }
        host->detect_change = 0;
 
        /* if there still is a card present, stop here */
drivers/mmc/core/core.h +1 -0
index 622085cd766f..5b48c50b5bb5 100644
@@ -78,6 +78,7 @@ void _mmc_detect_change(struct mmc_host *host, unsigned long delay,
 int _mmc_detect_card_removed(struct mmc_host *host);
 int mmc_detect_card_removed(struct mmc_host *host);
 
+int mmc_detect_mmc(struct mmc_host *host);
 int mmc_attach_mmc(struct mmc_host *host);
 int mmc_attach_sd(struct mmc_host *host);
 int mmc_attach_sdio(struct mmc_host *host);
drivers/mmc/core/mmc.c +0 -0
index 5be9b42d5057..e73de34e072b 100644
@@ -2316,6 +2316,14 @@ static const struct mmc_bus_ops mmc_ops = {
        .flush_cache = _mmc_flush_cache,
 };
 
Quoted 14 lines
Quoted 12 lines
drivers/mmc/core/mmc.c +8 -1
+/*
+ * Export function for mmc detection
+ */
+int mmc_detect_mmc(struct mmc_host *host){
+       mmc_detect(host);
+       return 0;
+}
+
 /*
  * Starting point for MMC card init.
  */
--
Quoted 2 lines

2.34.1

Hi Ulf

I worked On a PicoCalc board I added an extension board for the Raspberry Pi Zero 2W, while still wanting the Zero 2W to access the PicoCalc’s own SPI-connected SD card.

Here is the DTS I wrote:
```
/dts-v1/;
/plugin/;

/ {
    compatible = "brcm,bcm2835";

    fragment@0 {
        target = <&spidev0>;
        __overlay__ {
            status = "disabled";
        };
    };

    fragment@1 {
        target = <&spi1>;
        __overlay__ {
            status = "okay";

            #address-cells = <1>;
            #size-cells = <0>;
Quoted 12 lines
Quoted 11 lines
            sdslot@0 {
                compatible = "mmc-spi-slot";
                reg = <0>;                        // SPI CS0
                voltage-ranges = <3300 3300>;
                spi-max-frequency = <8000000>;
                gpios = <&gpio 6 1>;
                cd-gpios = <&gpio 6 1>;
            };
        };
    };
};
Quoted 1 line

```

Thanks for sharing this information!

Quoted 8 lines
Quoted 8 lines

With this setup, card insertion is detected, but removal is not. From what I can tell, mmc_sd_alive() itself is fine, but it seems does not check the state of cd-gpios.

For removable SPI SD cards that provide cd-gpios, I believe relying on the cd-gpios state may be a more direct and accurate way to determine card presence.

Sending a CMD13 to poll for the card should work as well, I think.

Or maybe there is something specific with SPI here?

Quoted 4 lines
Quoted 4 lines

Therefore, I added a check in _mmc_detect_card_removed() to handle the case of SPI-protocol cards with cd-gpios, so that removal detection can work properly.

I see. I am not sure (yet), that this is really the problem.

I had a look at the mmc_spi_probe() and there are problems with the order of how things are being registered. In general, mmc_gpiod_request_cd() should be called *before* calling mmc_add_host(), but I assume there are some legacy reasons as to why this isn't the case here.

That said, can you check if you are receiving card detect/removal interrupts and thus mmc_spi_detect_irq() should be invoked, when you insert/remove the card?

Does the card get detected if you start without a card in the slot when you boot and insert it later?

If you are not using the card detect interrupt, MMC_CAP_NEEDS_POLL must be set. Can you check if this is the case?

Quoted 3 lines
Quoted 3 lines

I am not deeply familiar with the Linux kernel, so if I am wrong please correct me. Thanks a lot.

No worries, I will try to help you as much as I can.

[...]

Kind regards
Uffe

author
#5 lore raw
Quoted 153 lines · Ulf Hansson
Quoted 131 lines · cuu
Quoted 129 lines · cuu
Quoted 85 lines · Ulf Hansson
Quoted 9 lines · cuu

Previously, SD cards using pure SPI protocol with mmc_spi could not reliably detect card removal via cd-gpios. This patch fixes the issue by extending the card alive check in _mmc_detect_card_removed() to include cd-gpios detection.

Additionally, mmc_rescan() now includes a status probing function specific to SPI SD cards to ensure the card state is correctly updated on insertion and removal events.

Sorry, I don't get this.

What is so special with SPI cards here that makes the regular card detection/removal procedures not to work?

Is there something wrong with mmc_sd_alive()?

Quoted 2 lines

Please use a proper name for your sob line. I assume "cuu" is an alias, no?

Kind regards
Uffe

Quoted 59 lines
---
 drivers/mmc/core/core.c | 15 ++++++++++++---
 drivers/mmc/core/core.h |  1 +
 drivers/mmc/core/mmc.c  |  8 ++++++++
 3 files changed, 21 insertions(+), 3 deletions(-)
drivers/mmc/core/core.c +12 -3
index 874c6fe92855..e028ec8fbf46 100644
@@ -2123,6 +2123,11 @@ int _mmc_detect_card_removed(struct mmc_host *host)
 
        ret = host->bus_ops->alive(host);
 
+        if(mmc_host_is_spi(host)) {
+               if(host->ops->get_cd){
+                               ret = !host->ops->get_cd(host);
+                }
+        }
        /*
         * Card detect status and alive check may be out of sync if card is
         * removed slowly, when card detect switch changes while card/slot
@@ -2238,9 +2243,13 @@ void mmc_rescan(struct work_struct *work)
        }
 
        /* Verify a registered card to be functional, else remove it. */
-       if (host->bus_ops)
-               host->bus_ops->detect(host);
-
+       if (host->bus_ops){
+               if (mmc_host_is_spi(host)){
+                       mmc_detect_mmc(host);
+               }else{
+                       host->bus_ops->detect(host);
+               }
+       }
        host->detect_change = 0;
 
        /* if there still is a card present, stop here */
drivers/mmc/core/core.h +1 -0
index 622085cd766f..5b48c50b5bb5 100644
@@ -78,6 +78,7 @@ void _mmc_detect_change(struct mmc_host *host, unsigned long delay,
 int _mmc_detect_card_removed(struct mmc_host *host);
 int mmc_detect_card_removed(struct mmc_host *host);
 
+int mmc_detect_mmc(struct mmc_host *host);
 int mmc_attach_mmc(struct mmc_host *host);
 int mmc_attach_sd(struct mmc_host *host);
 int mmc_attach_sdio(struct mmc_host *host);
drivers/mmc/core/mmc.c +0 -0
index 5be9b42d5057..e73de34e072b 100644
@@ -2316,6 +2316,14 @@ static const struct mmc_bus_ops mmc_ops = {
        .flush_cache = _mmc_flush_cache,
 };
 
Quoted 14 lines
Quoted 12 lines
drivers/mmc/core/mmc.c +8 -1
+/*
+ * Export function for mmc detection
+ */
+int mmc_detect_mmc(struct mmc_host *host){
+       mmc_detect(host);
+       return 0;
+}
+
 /*
  * Starting point for MMC card init.
  */
--
Quoted 2 lines

2.34.1

Hi Ulf

I worked On a PicoCalc board I added an extension board for the Raspberry Pi Zero 2W, while still wanting the Zero 2W to access the PicoCalc’s own SPI-connected SD card.

Here is the DTS I wrote:
```
/dts-v1/;
/plugin/;

/ {
    compatible = "brcm,bcm2835";

    fragment@0 {
        target = <&spidev0>;
        __overlay__ {
            status = "disabled";
        };
    };

    fragment@1 {
        target = <&spi1>;
        __overlay__ {
            status = "okay";

            #address-cells = <1>;
            #size-cells = <0>;
Quoted 12 lines
Quoted 11 lines
            sdslot@0 {
                compatible = "mmc-spi-slot";
                reg = <0>;                        // SPI CS0
                voltage-ranges = <3300 3300>;
                spi-max-frequency = <8000000>;
                gpios = <&gpio 6 1>;
                cd-gpios = <&gpio 6 1>;
            };
        };
    };
};
Quoted 1 line

```

Thanks for sharing this information!

Quoted 5 lines
Quoted 5 lines

With this setup, card insertion is detected, but removal is not. From what I can tell, mmc_sd_alive() itself is fine, but it seems does not check the state of cd-gpios.

Quoted 8 lines
Quoted 3 lines
Quoted 3 lines

For removable SPI SD cards that provide cd-gpios, I believe relying on the cd-gpios state may be a more direct and accurate way to determine card presence.

Sending a CMD13 to poll for the card should work as well, I think.

Or maybe there is something specific with SPI here?

There is nothing specific with SPI here.
I mentioned pure SPI because I don't think any SDIO high speed based sd card will have problems with linux mmc drivers since we used it for a long time.

But I am facing a simple SPI (4 wire, SDA SCL VCC GND) based card slot problem with detecting card removal.

I think CMD13 is not enough since we have cd-gpios here , I guess CMD13 is like to wait for it to be timeout so that the system will think the card is gone,right ?
but if we configured cd-gpios,then why wait for CMD13 ?

Quoted 5 lines
Quoted 4 lines
Quoted 4 lines

Therefore, I added a check in _mmc_detect_card_removed() to handle the case of SPI-protocol cards with cd-gpios, so that removal detection can work properly.

Quoted 7 lines

I see. I am not sure (yet), that this is really the problem.

I had a look at the mmc_spi_probe() and there are problems with the order of how things are being registered. In general, mmc_gpiod_request_cd() should be called *before* calling mmc_add_host(), but I assume there are some legacy reasons as to why this isn't the case here.

yeah, I guess so , so my first intention is not to break these legacy codes.

Quoted 3 lines

That said, can you check if you are receiving card detect/removal interrupts and thus mmc_spi_detect_irq() should be invoked, when you insert/remove the card?

In default , mmc_spi_detect_irq() won't be invoked here I've changed a little of my dts to force IRQ happen

```
/dts-v1/;
/plugin/;

/ {
    compatible = "brcm,bcm2835";

   fragment@0 {
                target = <&spidev0>;
                __overlay__ {
                        status = "disabled";
                };
   };

    fragment@1 {
        target = <&spi1>;
        __overlay__ {
            status = "okay";

            #address-cells = <1>;
            #size-cells = <0>;

            sdslot@0 {
                compatible = "mmc-spi-slot";
                reg = <0>;                        // SPI CS0
                voltage-ranges = <3300 3300>;
                spi-max-frequency = <8000000>;
                gpios = <&gpio 6 1>;
                cd-gpios = <&gpio 6 1>;
                interrupts-extended = <&gpio 6 1>;

            };
        };
    };
};
```

added one line interrupts-extended
so that we can see interrupts to invoke mmc_spi_detect_irq with different conditions
in dmesg | grep mmc ( I added a pr_info() in mmc_spi_detect_irq )

* interrupts-extended = <&gpio 6 2>;
insert invoked mmc_spi_detect_irq()
eject did not invoke mmc_spi_detect_irq()

* interrupts-extended = <&gpio 6 1>;
insert invoked mmc_spi_detect_irq()
eject the first time, won't invoke mmc_spi_detect_irq()
insert again and eject
now every time when inserting or ejecting, we can get a mmc_spi_detect_irq() invoked
It is a little bit weird,but that is what happened here
but still, linux did not remove the node /dev/mmcblk2( my spi sd card node)

* interrupts-extended = <&gpio 6 3>;
same as above

Quoted 2 lines

Does the card get detected if you start without a card in the slot when you boot and insert it later?

Yes, the sd card got detected when I inserted it after the system booted up.

Quoted 2 lines

If you are not using the card detect interrupt, MMC_CAP_NEEDS_POLL must be set. Can you check if this is the case?

By default MMC_CAP_NEEDS_POLL is not enabled here so I added one line

```
@@ -1281,6 +1282,8 @@ static int mmc_spi_probe(struct spi_device *spi)
                mmc->caps &= ~MMC_CAP_NEEDS_POLL;
                mmc_gpiod_request_cd_irq(mmc);
        }
+       mmc->caps |= MMC_CAP_NEEDS_POLL;
+
        mmc_detect_change(mmc, 0);

        /* Index 1 is write protect/read only */
```

in mmc_spi_probe() to force it to be enabled.
and tested it without IRQ modified dts

dmesg | grep mmc shows

```
[    9.186471] mmc_spi spi1.0: SD/MMC host mmc2, no WP, no poweroff, cd polling
[    9.234981] mmc2: host does not support reading read-only switch,
assuming write-enable
[    9.235039] mmc2: new SDHC card on SPI
[    9.238720] mmcblk2: mmc2:0000 LX32G 29.5 GiB
[    9.256975]  mmcblk2: p1 p2
[    9.263982] mmcblk2: mmc2:0000 LX32G 29.5 GiB (quirks 0x40000000)
```

as we can see , cd polling showed up

but polling is not working as we expected
nothing happened when I popped out my card, /dev/mmcblk2p* still remains in linux.
no more other lines in dmesg ,too

Quoted 10 lines
Quoted 3 lines
Quoted 3 lines

I am not deeply familiar with the Linux kernel, so if I am wrong please correct me. Thanks a lot.

No worries, I will try to help you as much as I can.

[...]

Kind regards
Uffe

All my testings are under rpi zero 2w with
Linux PicoCalc 6.17.0-rc6-v7+ #2 SMP Wed Sep 17 15:21:08 CST 2025 armv7l GNU/Linux

Kind regards
dexter kidd

#6 lore raw
Quoted 169 lines · dexter K
Quoted 153 lines · Ulf Hansson
Quoted 131 lines · cuu
Quoted 129 lines · cuu
Quoted 85 lines · Ulf Hansson
Quoted 9 lines · cuu

Previously, SD cards using pure SPI protocol with mmc_spi could not reliably detect card removal via cd-gpios. This patch fixes the issue by extending the card alive check in _mmc_detect_card_removed() to include cd-gpios detection.

Additionally, mmc_rescan() now includes a status probing function specific to SPI SD cards to ensure the card state is correctly updated on insertion and removal events.

Sorry, I don't get this.

What is so special with SPI cards here that makes the regular card detection/removal procedures not to work?

Is there something wrong with mmc_sd_alive()?

Quoted 2 lines

Please use a proper name for your sob line. I assume "cuu" is an alias, no?

Kind regards
Uffe

Quoted 59 lines
---
 drivers/mmc/core/core.c | 15 ++++++++++++---
 drivers/mmc/core/core.h |  1 +
 drivers/mmc/core/mmc.c  |  8 ++++++++
 3 files changed, 21 insertions(+), 3 deletions(-)
drivers/mmc/core/core.c +12 -3
index 874c6fe92855..e028ec8fbf46 100644
@@ -2123,6 +2123,11 @@ int _mmc_detect_card_removed(struct mmc_host *host)
 
        ret = host->bus_ops->alive(host);
 
+        if(mmc_host_is_spi(host)) {
+               if(host->ops->get_cd){
+                               ret = !host->ops->get_cd(host);
+                }
+        }
        /*
         * Card detect status and alive check may be out of sync if card is
         * removed slowly, when card detect switch changes while card/slot
@@ -2238,9 +2243,13 @@ void mmc_rescan(struct work_struct *work)
        }
 
        /* Verify a registered card to be functional, else remove it. */
-       if (host->bus_ops)
-               host->bus_ops->detect(host);
-
+       if (host->bus_ops){
+               if (mmc_host_is_spi(host)){
+                       mmc_detect_mmc(host);
+               }else{
+                       host->bus_ops->detect(host);
+               }
+       }
        host->detect_change = 0;
 
        /* if there still is a card present, stop here */
drivers/mmc/core/core.h +1 -0
index 622085cd766f..5b48c50b5bb5 100644
@@ -78,6 +78,7 @@ void _mmc_detect_change(struct mmc_host *host, unsigned long delay,
 int _mmc_detect_card_removed(struct mmc_host *host);
 int mmc_detect_card_removed(struct mmc_host *host);
 
+int mmc_detect_mmc(struct mmc_host *host);
 int mmc_attach_mmc(struct mmc_host *host);
 int mmc_attach_sd(struct mmc_host *host);
 int mmc_attach_sdio(struct mmc_host *host);
drivers/mmc/core/mmc.c +0 -0
index 5be9b42d5057..e73de34e072b 100644
@@ -2316,6 +2316,14 @@ static const struct mmc_bus_ops mmc_ops = {
        .flush_cache = _mmc_flush_cache,
 };
 
Quoted 14 lines
Quoted 12 lines
drivers/mmc/core/mmc.c +8 -1
+/*
+ * Export function for mmc detection
+ */
+int mmc_detect_mmc(struct mmc_host *host){
+       mmc_detect(host);
+       return 0;
+}
+
 /*
  * Starting point for MMC card init.
  */
--
Quoted 2 lines

2.34.1

Hi Ulf

I worked On a PicoCalc board I added an extension board for the Raspberry Pi Zero 2W, while still wanting the Zero 2W to access the PicoCalc’s own SPI-connected SD card.

Here is the DTS I wrote:
```
/dts-v1/;
/plugin/;

/ {
    compatible = "brcm,bcm2835";

    fragment@0 {
        target = <&spidev0>;
        __overlay__ {
            status = "disabled";
        };
    };

    fragment@1 {
        target = <&spi1>;
        __overlay__ {
            status = "okay";

            #address-cells = <1>;
            #size-cells = <0>;
Quoted 12 lines
Quoted 11 lines
            sdslot@0 {
                compatible = "mmc-spi-slot";
                reg = <0>;                        // SPI CS0
                voltage-ranges = <3300 3300>;
                spi-max-frequency = <8000000>;
                gpios = <&gpio 6 1>;
                cd-gpios = <&gpio 6 1>;
            };
        };
    };
};
Quoted 1 line

```

Thanks for sharing this information!

Quoted 5 lines
Quoted 5 lines

With this setup, card insertion is detected, but removal is not. From what I can tell, mmc_sd_alive() itself is fine, but it seems does not check the state of cd-gpios.

Quoted 8 lines
Quoted 3 lines
Quoted 3 lines

For removable SPI SD cards that provide cd-gpios, I believe relying on the cd-gpios state may be a more direct and accurate way to determine card presence.

Sending a CMD13 to poll for the card should work as well, I think.

Or maybe there is something specific with SPI here?

There is nothing specific with SPI here.
I mentioned pure SPI because I don't think any SDIO high speed based sd card will have problems with linux mmc drivers since we used it for a long time.

Quoted 6 lines

But I am facing a simple SPI (4 wire, SDA SCL VCC GND) based card slot problem with detecting card removal.

I think CMD13 is not enough since we have cd-gpios here , I guess CMD13 is like to wait for it to be timeout so that the system will think the card is gone,right ?

Correct!

Quoted 1 line

but if we configured cd-gpios,then why wait for CMD13 ?

CMD13 should be really quick, even in the timeout case we should only wait for a few us.

Of course, if there is a problem with the mmc host controller/driver, the timeout could be longer.

Quoted 67 lines
Quoted 6 lines
Quoted 5 lines
Quoted 5 lines

Therefore, I added a check in _mmc_detect_card_removed() to handle the case of SPI-protocol cards with cd-gpios, so that removal detection can work properly.

Quoted 7 lines

I see. I am not sure (yet), that this is really the problem.

I had a look at the mmc_spi_probe() and there are problems with the order of how things are being registered. In general, mmc_gpiod_request_cd() should be called *before* calling mmc_add_host(), but I assume there are some legacy reasons as to why this isn't the case here.

yeah, I guess so , so my first intention is not to break these legacy codes.

Quoted 3 lines

That said, can you check if you are receiving card detect/removal interrupts and thus mmc_spi_detect_irq() should be invoked, when you insert/remove the card?

In default , mmc_spi_detect_irq() won't be invoked here I've changed a little of my dts to force IRQ happen

```
/dts-v1/;
/plugin/;

/ {
    compatible = "brcm,bcm2835";

   fragment@0 {
                target = <&spidev0>;
                __overlay__ {
                        status = "disabled";
                };
   };

    fragment@1 {
        target = <&spi1>;
        __overlay__ {
            status = "okay";

            #address-cells = <1>;
            #size-cells = <0>;

            sdslot@0 {
                compatible = "mmc-spi-slot";
                reg = <0>;                        // SPI CS0
                voltage-ranges = <3300 3300>;
                spi-max-frequency = <8000000>;
                gpios = <&gpio 6 1>;
                cd-gpios = <&gpio 6 1>;
                interrupts-extended = <&gpio 6 1>;

            };
        };
    };
};
```

added one line interrupts-extended
so that we can see interrupts to invoke mmc_spi_detect_irq with different conditions
in dmesg | grep mmc ( I added a pr_info() in mmc_spi_detect_irq )

* interrupts-extended = <&gpio 6 2>;
insert invoked mmc_spi_detect_irq()
eject did not invoke mmc_spi_detect_irq()

Quoted 8 lines

* interrupts-extended = <&gpio 6 1>;
insert invoked mmc_spi_detect_irq()
eject the first time, won't invoke mmc_spi_detect_irq()
insert again and eject
now every time when inserting or ejecting, we can get a mmc_spi_detect_irq() invoked
It is a little bit weird,but that is what happened here
but still, linux did not remove the node /dev/mmcblk2( my spi sd card node)

Okay, so if we are not receiving an irq at card-removal this becomes a problem. If it's because of GPIO or pinctrl settings needs to be adjusted, I can't tell, because of limited HW knowledge in this case.

Nevertheless, if there are no irq available, we should fallback to using the polling mode with CMD13 (MMC_CAP_NEEDS_POLL).

That said, to me, it looks like we should start by trying to make the polling mode work first. If it doesn't, we need to understand why.

Quoted 16 lines

* interrupts-extended = <&gpio 6 3>;
same as above

Quoted 2 lines

Does the card get detected if you start without a card in the slot when you boot and insert it later?

Yes, the sd card got detected when I inserted it after the system booted up.

Quoted 2 lines

If you are not using the card detect interrupt, MMC_CAP_NEEDS_POLL must be set. Can you check if this is the case?

By default MMC_CAP_NEEDS_POLL is not enabled here so I added one line ```

Review static int mmc_spi_probe(struct spi_device *spi) · line 1286

Quoted 5 lines
(quoted diff) +2 -0
@@ -1281,6 +1282,8 @@ static int mmc_spi_probe(struct spi_device *spi)
                mmc->caps &= ~MMC_CAP_NEEDS_POLL;
                mmc_gpiod_request_cd_irq(mmc);
        }
+       mmc->caps |= MMC_CAP_NEEDS_POLL;
+

This needs to be set before calling mmc_add_host(). And beyond that, make sure we don't request for GPIO irq (don't call mmc_gpiod_request_cd_irq()), to make sure we only test polling mode separately.

Moreover, you may need to make changes in drivers/mmc/host/of_mmc_spi.c to avoid this too. Huh, this code is messy, sorry!

Quoted 16 lines
(quoted diff) +0 -0
        mmc_detect_change(mmc, 0);
 
        /* Index 1 is write protect/read only */
```
in  mmc_spi_probe() to force it to be enabled.
and tested it without IRQ modified dts

dmesg | grep mmc  shows
```
[    9.186471] mmc_spi spi1.0: SD/MMC host mmc2, no WP, no poweroff, cd polling
[    9.234981] mmc2: host does not support reading read-only switch,
assuming write-enable
[    9.235039] mmc2: new SDHC card on SPI
[    9.238720] mmcblk2: mmc2:0000 LX32G 29.5 GiB
[    9.256975]  mmcblk2: p1 p2
[    9.263982] mmcblk2: mmc2:0000 LX32G 29.5 GiB (quirks 0x40000000)
Quoted 7 lines

```
as we can see , cd polling showed up

but polling is not working as we expected
nothing happened when I popped out my card, /dev/mmcblk2p* still remains in linux.
no more other lines in dmesg ,too

Could it be that the CMD13 is hanging when there is no card inserted in the spi-slot?

Quoted 12 lines
Quoted 11 lines
Quoted 4 lines
Quoted 4 lines

I am not deeply familiar with the Linux kernel, so if I am wrong please correct me. Thanks a lot.

No worries, I will try to help you as much as I can.

[...]

Kind regards
Uffe

Quoted 7 lines

All my testings are under rpi zero 2w with
Linux PicoCalc 6.17.0-rc6-v7+ #2 SMP Wed Sep 17 15:21:08 CST 2025 armv7l GNU/Linux

Kind regards
dexter kidd

Apologize for the delay!

Kind regards
Uffe