[PATCH RFC] thunderbolt: Skip reading Rsvd SB registers on older-gen hardware

[PATCH RFC] thunderbolt: Skip reading Rsvd SB registers on older-gen hardware

[Konrad Dybcio]

From: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>

The Linux TBT subsystem supports hardware aligned with the latest
USB4 v2.0 specification. In some places though, it assumes registers
only defined in that specification version (previously marked as
Reserved) are always accessible.

Every version of the spec states:

"""
Unless specified otherwise, the Connection Manager shall not write a
register with a value that is marked as “Rsvd”. Writing a register with
a value that is marked as “Rsvd” results in undefined behavior.
"""

The effective list of SB registers that need this guarding currently
exclusively contains ones that aren't used outside the debugfs dump
logic, so just add the required checks there.

Fixes: 54e418106c76 ("thunderbolt: Add debugfs interface")
Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
---
 drivers/thunderbolt/debugfs.c | 20 ++++++++++++--------
 drivers/thunderbolt/retimer.c | 11 ++++++++++-
 drivers/thunderbolt/sb_regs.h | 11 ++++++-----
 drivers/thunderbolt/tb.h      |  2 ++
 4 files changed, 30 insertions(+), 14 deletions(-)

diff --git a/drivers/thunderbolt/debugfs.c b/drivers/thunderbolt/debugfs.c
index 042f6a0d0f7f..b74685bcd2a8 100644
--- a/drivers/thunderbolt/debugfs.c
+++ b/drivers/thunderbolt/debugfs.c
@@ -67,6 +67,7 @@ enum usb4_margin_cap_time_indp {
 struct sb_reg {
 	unsigned int reg;
 	unsigned int size;
+	u8 min_spec_ver;
 };
 
 #define SB_MAX_SIZE		64
@@ -75,9 +76,9 @@ struct sb_reg {
 static const struct sb_reg port_sb_regs[] = {
 	{ USB4_SB_VENDOR_ID, 4 },
 	{ USB4_SB_PRODUCT_ID, 4 },
-	{ USB4_SB_DEBUG_CONF, 4 },
-	{ USB4_SB_DEBUG, 54 },
-	{ USB4_SB_LRD_TUNING, 4 },
+	{ USB4_SB_DEBUG_CONF, 4, .min_spec_ver = 2 },
+	{ USB4_SB_DEBUG, 54, .min_spec_ver = 2 },
+	{ USB4_SB_LRD_TUNING, 4, .min_spec_ver = 2 },
 	{ USB4_SB_OPCODE, 4 },
 	{ USB4_SB_METADATA, 4 },
 	{ USB4_SB_LINK_CONF, 3 },
@@ -92,11 +93,11 @@ static const struct sb_reg retimer_sb_regs[] = {
 	{ USB4_SB_VENDOR_ID, 4 },
 	{ USB4_SB_PRODUCT_ID, 4 },
 	{ USB4_SB_FW_VERSION, 4 },
-	{ USB4_SB_LRD_TUNING, 4 },
+	{ USB4_SB_LRD_TUNING, 4, .min_spec_ver = 2 },
 	{ USB4_SB_OPCODE, 4 },
 	{ USB4_SB_METADATA, 4 },
 	{ USB4_SB_GEN23_TXFFE, 4 },
-	{ USB4_SB_GEN4_TXFFE, 4 },
+	{ USB4_SB_GEN4_TXFFE, 4, .min_spec_ver = 2 },
 	{ USB4_SB_VERSION, 4 },
 	{ USB4_SB_DATA, 64 },
 };
@@ -2347,7 +2348,7 @@ DEBUGFS_ATTR_RW(counters);
 
 static int sb_regs_show(struct tb_port *port, const struct sb_reg *sb_regs,
 			size_t size, enum usb4_sb_target target, u8 index,
-			struct seq_file *s)
+			u8 spec_version, struct seq_file *s)
 {
 	int ret, i;
 
@@ -2358,6 +2359,9 @@ static int sb_regs_show(struct tb_port *port, const struct sb_reg *sb_regs,
 		u8 data[64];
 		int j;
 
+		if (regs->min_spec_ver > spec_version)
+			continue;
+
 		memset(data, 0, sizeof(data));
 		ret = usb4_port_sb_read(port, target, index, regs->reg, data,
 					regs->size);
@@ -2388,7 +2392,7 @@ static int port_sb_regs_show(struct seq_file *s, void *not_used)
 	}
 
 	ret = sb_regs_show(port, port_sb_regs, ARRAY_SIZE(port_sb_regs),
-			   USB4_SB_TARGET_ROUTER, 0, s);
+			   USB4_SB_TARGET_ROUTER, 0, port->config.thunderbolt_version, s);
 
 	mutex_unlock(&tb->lock);
 out_rpm_put:
@@ -2503,7 +2507,7 @@ static int retimer_sb_regs_show(struct seq_file *s, void *not_used)
 	}
 
 	ret = sb_regs_show(rt->port, retimer_sb_regs, ARRAY_SIZE(retimer_sb_regs),
-			   USB4_SB_TARGET_RETIMER, rt->index, s);
+			   USB4_SB_TARGET_RETIMER, rt->index, rt->spec_version, s);
 
 	mutex_unlock(&tb->lock);
 out_rpm_put:
diff --git a/drivers/thunderbolt/retimer.c b/drivers/thunderbolt/retimer.c
index b1e8d9d6454d..b27a34fe83ea 100644
--- a/drivers/thunderbolt/retimer.c
+++ b/drivers/thunderbolt/retimer.c
@@ -389,8 +389,8 @@ const struct device_type tb_retimer_type = {
 static int tb_retimer_add(struct tb_port *port, u8 index, u32 auth_status,
 			  bool on_board)
 {
+	u32 vendor, device, version;
 	struct tb_retimer *rt;
-	u32 vendor, device;
 	int ret;
 
 	ret = usb4_port_sb_read(port, USB4_SB_TARGET_RETIMER, index,
@@ -409,6 +409,14 @@ static int tb_retimer_add(struct tb_port *port, u8 index, u32 auth_status,
 		return ret;
 	}
 
+	ret = usb4_port_sb_read(port, USB4_SB_TARGET_RETIMER, index,
+				USB4_SB_FW_VERSION, &version, sizeof(version));
+	if (ret) {
+		if (ret != -ENODEV)
+			tb_port_warn(port, "failed read retimer Version: %d\n", ret);
+		return ret;
+	}
+
 
 	rt = kzalloc_obj(*rt);
 	if (!rt)
@@ -417,6 +425,7 @@ static int tb_retimer_add(struct tb_port *port, u8 index, u32 auth_status,
 	rt->index = index;
 	rt->vendor = vendor;
 	rt->device = device;
+	rt->spec_version = FIELD_GET(USB4_SB_FW_VERSION_VERSION, version);
 	rt->auth_status = auth_status;
 	rt->port = port;
 	rt->tb = port->sw->tb;
diff --git a/drivers/thunderbolt/sb_regs.h b/drivers/thunderbolt/sb_regs.h
index 5391502a4b87..a9c2d22ff69f 100644
--- a/drivers/thunderbolt/sb_regs.h
+++ b/drivers/thunderbolt/sb_regs.h
@@ -12,10 +12,11 @@
 
 #define USB4_SB_VENDOR_ID			0x00
 #define USB4_SB_PRODUCT_ID			0x01
-#define USB4_SB_FW_VERSION			0x02
-#define USB4_SB_DEBUG_CONF			0x05
-#define USB4_SB_DEBUG				0x06
-#define USB4_SB_LRD_TUNING			0x07
+#define USB4_SB_FW_VERSION			0x02 /* Retimer-only */
+#define USB4_SB_FW_VERSION_VERSION		GENMASK(7, 0)
+#define USB4_SB_DEBUG_CONF			0x05 /* Port-only, USB4 v2.0 */
+#define USB4_SB_DEBUG				0x06 /* Port-only, USB4 v2.0 */
+#define USB4_SB_LRD_TUNING			0x07 /* USB4 v2.0 */
 #define USB4_SB_OPCODE				0x08
 
 enum usb4_sb_opcode {
@@ -42,7 +43,7 @@ enum usb4_sb_opcode {
 #define USB4_SB_METADATA_NVM_AUTH_WRITE_MASK	GENMASK(5, 0)
 #define USB4_SB_LINK_CONF			0x0c
 #define USB4_SB_GEN23_TXFFE			0x0d
-#define USB4_SB_GEN4_TXFFE			0x0e
+#define USB4_SB_GEN4_TXFFE			0x0e /* USB4 v2.0 */
 #define USB4_SB_VERSION				0x0f
 #define USB4_SB_DATA				0x12
 
diff --git a/drivers/thunderbolt/tb.h b/drivers/thunderbolt/tb.h
index 217c3114bec8..7eba058e9c2b 100644
--- a/drivers/thunderbolt/tb.h
+++ b/drivers/thunderbolt/tb.h
@@ -335,6 +335,7 @@ struct usb4_port {
  * @no_nvm_upgrade: Prevent NVM upgrade of this retimer
  * @auth_status: Status of last NVM authentication
  * @margining: Pointer to margining structure if enabled
+ * @spec_version: The implemented USB4 Retimer Specification version
  */
 struct tb_retimer {
 	struct device dev;
@@ -346,6 +347,7 @@ struct tb_retimer {
 	struct tb_nvm *nvm;
 	bool no_nvm_upgrade;
 	u32 auth_status;
+	u8 spec_version;
 #ifdef CONFIG_USB4_DEBUGFS_MARGINING
 	struct tb_margining *margining;
 #endif

---
base-commit: db7efce4ae23ad5e42f5f55428f529ff62b86fab
change-id: 20260413-topic-usb4_limit_sb_reads-965ed26524a1

Best regards,
-- 
Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>

Re: [PATCH RFC] thunderbolt: Skip reading Rsvd SB registers on older-gen hardware

[Greg Kroah-Hartman]

On Mon, Apr 13, 2026 at 12:54:41PM +0200, Konrad Dybcio wrote:
> From: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> 
> The Linux TBT subsystem supports hardware aligned with the latest
> USB4 v2.0 specification. In some places though, it assumes registers
> only defined in that specification version (previously marked as
> Reserved) are always accessible.
> 
> Every version of the spec states:
> 
> """
> Unless specified otherwise, the Connection Manager shall not write a
> register with a value that is marked as “Rsvd”. Writing a register with
> a value that is marked as “Rsvd” results in undefined behavior.
> """
> 
> The effective list of SB registers that need this guarding currently
> exclusively contains ones that aren't used outside the debugfs dump
> logic, so just add the required checks there.
> 
> Fixes: 54e418106c76 ("thunderbolt: Add debugfs interface")
> Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> ---
>  drivers/thunderbolt/debugfs.c | 20 ++++++++++++--------
>  drivers/thunderbolt/retimer.c | 11 ++++++++++-
>  drivers/thunderbolt/sb_regs.h | 11 ++++++-----
>  drivers/thunderbolt/tb.h      |  2 ++
>  4 files changed, 30 insertions(+), 14 deletions(-)
> 

Hi,

This is the friendly patch-bot of Greg Kroah-Hartman.  You have sent him
a patch that has triggered this response.  He used to manually respond
to these common problems, but in order to save his sanity (he kept
writing the same thing over and over, yet to different people), I was
created.  Hopefully you will not take offence and will fix the problem
in your patch and resubmit it so that it can be accepted into the Linux
kernel tree.

You are receiving this message because of the following common error(s)
as indicated below:

- You have marked a patch with a "Fixes:" tag for a commit that is in an
  older released kernel, yet you do not have a cc: stable line in the
  signed-off-by area at all, which means that the patch will not be
  applied to any older kernel releases.  To properly fix this, please
  follow the documented rules in the
  Documentation/process/stable-kernel-rules.rst file for how to resolve
  this.

If you wish to discuss this problem further, or you have questions about
how to resolve this issue, please feel free to respond to this email and
Greg will reply once he has dug out from the pending patches received
from other developers.

thanks,

greg k-h's patch email bot

Re: [PATCH RFC] thunderbolt: Skip reading Rsvd SB registers on older-gen hardware

[Mika Westerberg]

On Mon, Apr 13, 2026 at 12:54:41PM +0200, Konrad Dybcio wrote:
> From: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> 
> The Linux TBT subsystem supports hardware aligned with the latest
> USB4 v2.0 specification. In some places though, it assumes registers
> only defined in that specification version (previously marked as
> Reserved) are always accessible.
> 
> Every version of the spec states:
> 
> """
> Unless specified otherwise, the Connection Manager shall not write a
> register with a value that is marked as “Rsvd”. Writing a register with
> a value that is marked as “Rsvd” results in undefined behavior.
> """
> 
> The effective list of SB registers that need this guarding currently
> exclusively contains ones that aren't used outside the debugfs dump
> logic, so just add the required checks there.
> 
> Fixes: 54e418106c76 ("thunderbolt: Add debugfs interface")
> Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> ---
>  drivers/thunderbolt/debugfs.c | 20 ++++++++++++--------
>  drivers/thunderbolt/retimer.c | 11 ++++++++++-
>  drivers/thunderbolt/sb_regs.h | 11 ++++++-----
>  drivers/thunderbolt/tb.h      |  2 ++
>  4 files changed, 30 insertions(+), 14 deletions(-)

This is alternative for the v2 patch you sent earlier, right? I prefer that
one over this.

Re: [PATCH RFC] thunderbolt: Skip reading Rsvd SB registers on older-gen hardware

[Konrad Dybcio]

On 4/13/26 1:40 PM, Mika Westerberg wrote:
> On Mon, Apr 13, 2026 at 12:54:41PM +0200, Konrad Dybcio wrote:
>> From: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
>>
>> The Linux TBT subsystem supports hardware aligned with the latest
>> USB4 v2.0 specification. In some places though, it assumes registers
>> only defined in that specification version (previously marked as
>> Reserved) are always accessible.
>>
>> Every version of the spec states:
>>
>> """
>> Unless specified otherwise, the Connection Manager shall not write a
>> register with a value that is marked as “Rsvd”. Writing a register with
>> a value that is marked as “Rsvd” results in undefined behavior.
>> """
>>
>> The effective list of SB registers that need this guarding currently
>> exclusively contains ones that aren't used outside the debugfs dump
>> logic, so just add the required checks there.
>>
>> Fixes: 54e418106c76 ("thunderbolt: Add debugfs interface")
>> Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
>> ---
>>  drivers/thunderbolt/debugfs.c | 20 ++++++++++++--------
>>  drivers/thunderbolt/retimer.c | 11 ++++++++++-
>>  drivers/thunderbolt/sb_regs.h | 11 ++++++-----
>>  drivers/thunderbolt/tb.h      |  2 ++
>>  4 files changed, 30 insertions(+), 14 deletions(-)
> 
> This is alternative for the v2 patch you sent earlier, right? I prefer that
> one over this.

I think they're complementary. This patch ensures compliance with the
quoted part of the spec, while the other one improves the UX and aids
debugging.

Konrad

Re: [PATCH RFC] thunderbolt: Skip reading Rsvd SB registers on older-gen hardware

[Mika Westerberg]

On Mon, Apr 13, 2026 at 01:43:49PM +0200, Konrad Dybcio wrote:
> On 4/13/26 1:40 PM, Mika Westerberg wrote:
> > On Mon, Apr 13, 2026 at 12:54:41PM +0200, Konrad Dybcio wrote:
> >> From: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> >>
> >> The Linux TBT subsystem supports hardware aligned with the latest
> >> USB4 v2.0 specification. In some places though, it assumes registers
> >> only defined in that specification version (previously marked as
> >> Reserved) are always accessible.
> >>
> >> Every version of the spec states:
> >>
> >> """
> >> Unless specified otherwise, the Connection Manager shall not write a
> >> register with a value that is marked as “Rsvd”. Writing a register with
> >> a value that is marked as “Rsvd” results in undefined behavior.
> >> """
> >>
> >> The effective list of SB registers that need this guarding currently
> >> exclusively contains ones that aren't used outside the debugfs dump
> >> logic, so just add the required checks there.
> >>
> >> Fixes: 54e418106c76 ("thunderbolt: Add debugfs interface")
> >> Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
> >> ---
> >>  drivers/thunderbolt/debugfs.c | 20 ++++++++++++--------
> >>  drivers/thunderbolt/retimer.c | 11 ++++++++++-
> >>  drivers/thunderbolt/sb_regs.h | 11 ++++++-----
> >>  drivers/thunderbolt/tb.h      |  2 ++
> >>  4 files changed, 30 insertions(+), 14 deletions(-)
> > 
> > This is alternative for the v2 patch you sent earlier, right? I prefer that
> > one over this.
> 
> I think they're complementary. This patch ensures compliance with the
> quoted part of the spec, while the other one improves the UX and aids
> debugging.

This adds a lot of code for just debugfs access so I think we are better
without.

Re: [PATCH RFC] thunderbolt: Skip reading Rsvd SB registers on older-gen hardware

[Konrad Dybcio]

On 4/13/26 1:53 PM, Mika Westerberg wrote:
> On Mon, Apr 13, 2026 at 01:43:49PM +0200, Konrad Dybcio wrote:
>> On 4/13/26 1:40 PM, Mika Westerberg wrote:
>>> On Mon, Apr 13, 2026 at 12:54:41PM +0200, Konrad Dybcio wrote:
>>>> From: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
>>>>
>>>> The Linux TBT subsystem supports hardware aligned with the latest
>>>> USB4 v2.0 specification. In some places though, it assumes registers
>>>> only defined in that specification version (previously marked as
>>>> Reserved) are always accessible.
>>>>
>>>> Every version of the spec states:
>>>>
>>>> """
>>>> Unless specified otherwise, the Connection Manager shall not write a
>>>> register with a value that is marked as “Rsvd”. Writing a register with
>>>> a value that is marked as “Rsvd” results in undefined behavior.
>>>> """
>>>>
>>>> The effective list of SB registers that need this guarding currently
>>>> exclusively contains ones that aren't used outside the debugfs dump
>>>> logic, so just add the required checks there.
>>>>
>>>> Fixes: 54e418106c76 ("thunderbolt: Add debugfs interface")
>>>> Signed-off-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
>>>> ---
>>>>  drivers/thunderbolt/debugfs.c | 20 ++++++++++++--------
>>>>  drivers/thunderbolt/retimer.c | 11 ++++++++++-
>>>>  drivers/thunderbolt/sb_regs.h | 11 ++++++-----
>>>>  drivers/thunderbolt/tb.h      |  2 ++
>>>>  4 files changed, 30 insertions(+), 14 deletions(-)
>>>
>>> This is alternative for the v2 patch you sent earlier, right? I prefer that
>>> one over this.
>>
>> I think they're complementary. This patch ensures compliance with the
>> quoted part of the spec, while the other one improves the UX and aids
>> debugging.
> 
> This adds a lot of code for just debugfs access so I think we are better
> without.

Just the other one works too, then.

Konrad