usb: composite: fix possible alignment issues
PATCH v2 1/2 usb: composite: fix possible alignment issues
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"
Fix this by converting two functions to use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.
---
Changes in v2:
- fix compiler warning "dereferencing ‘void *'"
drivers/usb/gadget/composite.c | 25 ++++++++++++++++++------- 1 file changed, 18 insertions(+), 7 deletions(-)
drivers/usb/gadget/composite.c +18 -7
@@ -12,8 +12,16 @@#define USB_BUFSIZ 4096 +/* Helper type for accessing packed u16 pointers */ +typedef struct { __le16 val; } __packed __le16_packed; +static struct usb_composite_driver *composite; +static inline void le16_add_cpu_packed (__le16_packed *var,u16 val) +{ + var->val =cpu_to_le16 (le16_to_cpu (var->val) + val); +} +/** * usb_add_function() - add a function to a configuration * @config: the configuration
@@ -480,20 +488,21 @@ done:* the host side. */ -static void collect_langs (struct usb_gadget_strings **sp,__le16 *buf) +static void collect_langs (struct usb_gadget_strings **sp,void *buf) {const struct usb_gadget_strings *s;u16 language; -__le16 *tmp; +__le16_packed *tmp; +__le16_packed *end = (buf +252 );while (*sp) { s = *sp; language =cpu_to_le16 (s->language); -for (tmp = buf; *tmp && tmp < &buf[126 ]; tmp++) { -if (*tmp == language) +for (tmp = buf; tmp->val && tmp < end; tmp++) { +if (tmp->val == language) goto repeat; } - *tmp++ = language; + tmp->val = language; repeat: sp++; }
@@ -705,7 +714,8 @@ static int bos_desc(struct usb_composite_dev *cdev)*/ usb_ext = cdev->req->buf +le16_to_cpu (bos->wTotalLength); bos->bNumDeviceCaps++; -le16_add_cpu (&bos->wTotalLength,USB_DT_USB_EXT_CAP_SIZE ); +le16_add_cpu_packed ((__le16_packed *)&bos->wTotalLength, +USB_DT_USB_EXT_CAP_SIZE ); usb_ext->bLength =USB_DT_USB_EXT_CAP_SIZE ; usb_ext->bDescriptorType =USB_DT_DEVICE_CAPABILITY ; usb_ext->bDevCapabilityType =USB_CAP_TYPE_EXT ;
@@ -721,7 +731,8 @@ static int bos_desc(struct usb_composite_dev *cdev) ss_cap = cdev->req->buf +le16_to_cpu (bos->wTotalLength); bos->bNumDeviceCaps++; -le16_add_cpu (&bos->wTotalLength,USB_DT_USB_SS_CAP_SIZE ); +le16_add_cpu_packed ((__le16_packed *)&bos->wTotalLength, +USB_DT_USB_SS_CAP_SIZE ); ss_cap->bLength =USB_DT_USB_SS_CAP_SIZE ; ss_cap->bDescriptorType =USB_DT_DEVICE_CAPABILITY ; ss_cap->bDevCapabilityType =USB_SS_CAP_TYPE ;
Signature · 1 line
2.20.1
PATCH v2 2/2 usb: dwc2: fix possible alignment issues
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"
Fix this by converting dwc2_fifo_read to use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.
---
Changes in v2: None
drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c +2 -2
@@ -731,7 +731,7 @@ static int write_fifo_ep0(struct dwc2_ep *ep, struct dwc2_request *req)return 0 ; } -static int dwc2_fifo_read (struct dwc2_ep *ep,u32 *cp,int max) +static int dwc2_fifo_read (struct dwc2_ep *ep,void *cp,int max) {invalidate_dcache_range ((unsigned long )cp, (unsigned long )cp +ROUND (max,CONFIG_SYS_CACHELINE_SIZE ));
@@ -1285,7 +1285,7 @@ static void dwc2_ep0_setup(struct dwc2_udc *dev)nuke (ep, -EPROTO );/* read control req from fifo (8 bytes) */ -dwc2_fifo_read (ep, (u32 *)usb_ctrl,8 ); +dwc2_fifo_read (ep, usb_ctrl,8 );debug_cond (DEBUG_SETUP !=0 , "%s: bRequestType = 0x%x(%s), bRequest = 0x%x"
Signature · 1 line
2.20.1
PATCH v2 2/2 usb: dwc2: fix possible alignment issues
Quoted 30 lines · On Thu, 21 Nov 2019 22:15:23 +0100 Simon Goldschmidt…
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Fix this by converting
dwc2_fifo_readto use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.Signed-off-by Simon Goldschmidt <[email protected]>---
Changes in v2: None
drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)(quoted diff) +0
-0
b/drivers/usb/gadget/dwc2_udc_otg_xfer_dma.cindex7eb632d3b1..dba221dad0100644 ---a/drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c+++b/drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c@@ -731,7 +731,7 @@ static intwrite_fifo_ep0(struct dwc2_ep *ep, struct dwc2_request *req)return 0; }drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c +2
-1-static int dwc2_fifo_read (struct dwc2_ep *ep,u32 *cp,int max) +static int dwc2_fifo_read (struct dwc2_ep *ep,void *cp,int max) { invalidate_dcache_range((unsigned long)cp, (unsigned long)cp +ROUND (max,CONFIG_SYS_CACHELINE_SIZE ));
drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c static void dwc2_ep0_setup(struct dwc2_udc *dev) · line 1291
Quoted 8 lines
drivers/usb/gadget/dwc2_udc_otg_xfer_dma.c +1
-1@@ -1285,7 +1285,7 @@ static void dwc2_ep0_setup(struct dwc2_udc *dev)nuke (ep, -EPROTO );/* read control req from fifo (8 bytes) */ -dwc2_fifo_read (ep, (u32 *)usb_ctrl,8 ); +dwc2_fifo_read (ep, usb_ctrl,8 );debug_cond (DEBUG_SETUP !=0 , "%s: bRequestType = 0x%x(%s), bRequest = 0x%x"
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20191121/a089ca0f/attachment.sig>
PATCH v2 1/2 usb: composite: fix possible alignment issues
In reply to Simon GoldschmidtQuoted 85 lines · On Thu, 21 Nov 2019 22:15:22 +0100 Simon Goldschmidt…
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Fix this by converting two functions to use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.
Signed-off-by Simon Goldschmidt <[email protected]>---
Changes in v2:
- fix compiler warning "dereferencing ‘void *'"drivers/usb/gadget/composite.c | 25 ++++++++++++++++++------- 1 file changed, 18 insertions(+), 7 deletions(-)(quoted diff) +0
-0
b/drivers/usb/gadget/composite.cindex618a7d5016..c98a444245100644drivers/usb/gadget/composite.c +17
-7@@ -12,8 +12,16 @@#define USB_BUFSIZ 4096 +/* Helper type for accessing packed u16 pointers */ +typedef struct { __le16 val; } __packed __le16_packed; +static struct usb_composite_driver *composite; +static inline void le16_add_cpu_packed (__le16_packed *var,u16 val) +{ + var->val =cpu_to_le16 (le16_to_cpu (var->val) + val); +} +/** * usb_add_function() - add a function to a configuration * @config: the configuration @@ -480,20 +488,21 @@ done:* the host side. */ -static void collect_langs (struct usb_gadget_strings **sp,__le16 *buf) +static void collect_langs(struct usb_gadget_strings **sp, void *buf) {const struct usb_gadget_strings *s;u16 language; -__le16 *tmp; +__le16_packed *tmp; +__le16_packed *end = (buf +252 );while (*sp) { s = *sp; language =cpu_to_le16 (s->language); -for (tmp = buf; *tmp && tmp < &buf[126 ]; tmp++) { -if (*tmp == language) +for (tmp = buf; tmp->val && tmp < end; tmp++) { +if (tmp->val == language) goto repeat; } - *tmp++ = language; + tmp->val = language; repeat: sp++; }@@ -705,7 +714,8 @@ static int bos_desc(struct usb_composite_dev*cdev) */ usb_ext = cdev->req->buf +le16_to_cpu (bos->wTotalLength); bos->bNumDeviceCaps++; -le16_add_cpu (&bos->wTotalLength,USB_DT_USB_EXT_CAP_SIZE ); +le16_add_cpu_packed ((__le16_packed *)&bos->wTotalLength, +USB_DT_USB_EXT_CAP_SIZE ); usb_ext->bLength =USB_DT_USB_EXT_CAP_SIZE ; usb_ext->bDescriptorType =USB_DT_DEVICE_CAPABILITY ; usb_ext->bDevCapabilityType =USB_CAP_TYPE_EXT ;@@ -721,7 +731,8 @@ static int bos_desc(struct usb_composite_dev *cdev) ss_cap = cdev->req->buf +le16_to_cpu (bos->wTotalLength); bos->bNumDeviceCaps++; -le16_add_cpu (&bos->wTotalLength, USB_DT_USB_SS_CAP_SIZE); +le16_add_cpu_packed ((__le16_packed *)&bos->wTotalLength, +USB_DT_USB_SS_CAP_SIZE ); ss_cap->bLength =USB_DT_USB_SS_CAP_SIZE ;
drivers/usb/gadget/composite.c static int bos_desc(struct usb_composite_dev *cdev) · line 738
Quoted 2 lines
drivers/usb/gadget/composite.c +0
-0ss_cap->bDescriptorType =USB_DT_DEVICE_CAPABILITY ; ss_cap->bDevCapabilityType =USB_SS_CAP_TYPE ;
Best regards,
Lukasz Majewski
--
DENX Software Engineering GmbH, Managing Director: Wolfgang Denk
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-59 Fax: (+49)-8142-66989-80 Email: lukma at denx.de -------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://lists.denx.de/pipermail/u-boot/attachments/20191121/28c4c32c/attachment.sig>
Marek Vasut
AppliedPATCH v2 1/2 usb: composite: fix possible alignment issues
In reply to Simon GoldschmidtQuoted 8 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Fix this by converting two functions to use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.
Signed-off-by Simon Goldschmidt <[email protected]>
Applied both, thanks.
PATCH v2 1/2 usb: composite: fix possible alignment issues
Quoted 5 lines · Marek Vasut
Quoted 4 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"
Quoted 7 lines
Quoted 4 lines
Fix this by converting two functions to use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.
Signed-off-by Simon Goldschmidt <[email protected]>Applied both, thanks.
With these two patches applied to origin/master GCC 9.2.1 does not produce warnings anymore but for tbs2910_defconfig:
u-boot.imx exceeds file size limit:
limit: 0x5fc00 bytes
actual: 0x60c00 bytes
excess: 0x1000 bytes
make: *** [Makefile:1159: u-boot.imx] Error 1
make: *** Deleting file 'u-boot.imx'
So irrespective of my patches for the USB keyboard we need to address the size limit on TBS2910.
Best regards
Heinrich
PATCH v2 1/2 usb: composite: fix possible alignment issues
Quoted 19 lines · Heinrich Schuchardt
Quoted 5 lines · Marek Vasut
Quoted 4 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Quoted 7 lines
Quoted 4 lines
Fix this by converting two functions to use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.
Signed-off-by Simon Goldschmidt <[email protected]>Applied both, thanks.
With these two patches applied to origin/master GCC 9.2.1 does not produce warnings anymore but for
tbs2910_defconfig:u-boot.imx exceeds file size limit:
Quoted 8 lines
limit: 0x5fc00 bytes actual: 0x60c00 bytes excess: 0x1000 bytesmake: *** [Makefile:1159: u-boot.imx] Error 1
make: *** Deleting file 'u-boot.imx'So irrespective of my patches for the USB keyboard we need to address the size limit on TBS2910.
Is that due to my cleanup patches? Can you compare the size by compiling without them? That should work if you leave away the -Werror switch.
Regards,
Simon
Quoted 4 lines
Best regards
Heinrich
PATCH v2 1/2 usb: composite: fix possible alignment issues
Quoted 26 lines · Simon Goldschmidt
Quoted 25 lines · Heinrich Schuchardt
Quoted 5 lines · Marek Vasut
Quoted 4 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Quoted 7 lines
Quoted 4 lines
Fix this by converting two functions to use unaligned access since packed structures may be on an unaligned address, depending on USB hardware.
Signed-off-by Simon Goldschmidt <[email protected]>Applied both, thanks.
With these two patches applied to origin/master GCC 9.2.1 does not produce warnings anymore but for
tbs2910_defconfig:u-boot.imx exceeds file size limit:
limit: 0x5fc00 bytes actual: 0x60c00 bytes excess: 0x1000 bytesmake: *** [Makefile:1159: u-boot.imx] Error 1
make: *** Deleting file 'u-boot.imx'
Quoted 8 lines
Quoted 2 lines
So irrespective of my patches for the USB keyboard we need to address the size limit on TBS2910.
Is that due to my cleanup patches? Can you compare the size by compiling without them? That should work if you leave away the -Werror switch.
Regards,
Simon
GCC 9.2.1 without your patches but with -Wno-address-of-packed-member:
u-boot.imx exceeds file size limit:
limit: 0x5fc00 bytes
actual: 0x60c00 bytes
excess: 0x1000 bytes
Best regards
Heinrich
Quoted 6 lines
Quoted 5 lines
Best regards
Heinrich
PATCH v2 1/2 usb: composite: fix possible alignment issues
Quoted 38 lines · Heinrich Schuchardt
Quoted 28 lines · Simon Goldschmidt
Quoted 26 lines · Heinrich Schuchardt
Quoted 5 lines · Marek Vasut
Quoted 4 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Quoted 8 lines
Quoted 5 lines
Fix this by converting two functions to use unaligned access since packed
structures may be on an unaligned address, depending on USB hardware.Signed-off-by Simon Goldschmidt <[email protected]>Applied both, thanks.
With these two patches applied to origin/master GCC 9.2.1 does not produce warnings anymore but for
tbs2910_defconfig:u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytes
make: *** [Makefile:1159: u-boot.imx] Error 1
make: *** Deleting file 'u-boot.imx'Quoted 8 lines
Quoted 2 lines
So irrespective of my patches for the USB keyboard we need to address the size limit on TBS2910.
Is that due to my cleanup patches? Can you compare the size by compiling without them? That should work if you leave away the -Werror switch.
Regards,
Simon
Quoted 6 lines
GCC 9.2.1 without your patches but with -Wno-address-of-packed-member:
u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytes
I see, so you have additional options added to the build which trigger the size issue. It would be nice to mention that up front. Do you use any other extra options ?
PATCH v2 1/2 usb: composite: fix possible alignment issues
Quoted 41 lines · Marek Vasut
Quoted 40 lines · Heinrich Schuchardt
Quoted 28 lines · Simon Goldschmidt
Quoted 26 lines · Heinrich Schuchardt
Quoted 5 lines · Marek Vasut
Quoted 4 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Quoted 8 lines
Quoted 5 lines
Fix this by converting two functions to use unaligned access since packed
structures may be on an unaligned address, depending on USB hardware.Signed-off-by Simon Goldschmidt <[email protected]>Applied both, thanks.
With these two patches applied to origin/master GCC 9.2.1 does not produce warnings anymore but for
tbs2910_defconfig:u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytes
make: *** [Makefile:1159: u-boot.imx] Error 1
make: *** Deleting file 'u-boot.imx'Quoted 8 lines
Quoted 2 lines
So irrespective of my patches for the USB keyboard we need to address the size limit on TBS2910.
Is that due to my cleanup patches? Can you compare the size by compiling without them? That should work if you leave away the -Werror switch.
Regards,
SimonGCC 9.2.1 without your patches but with -Wno-address-of-packed-member:
Quoted 9 lines
Quoted 4 lines
u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytesI see, so you have additional options added to the build which trigger the size issue. It would be nice to mention that up front. Do you use any other extra options ?
Dear Marek,
Simon asked me to determine if origin/master exceeds the u-boot.imx size limit when compiled without his patches. The only way to do so is to suppress the build warnings.
-Wno-address-of-packed-member is the only option I added to origin/master. This option suppresses the build error that we get without Simon's patches.
The only other difference between Gitlab CI and my setup is that I use GCC 9.2.1 (arm-linux-gnueabihf-gcc on Debian Bullseye).
These are the sizes of u-boot.bin:
390080 bytes with Simon's patches
390080 bytes with Simon's patches + -Wno-address-of-packed-member 390064 bytes origin/master + -Wno-address-of-packed-member 386248 bytes with Simon's patches + CONFIG_REGEX=n 390024 bytes with Simon patches +
"usb: kbd: simplify coding for arrow keys"
390440 bytes with Simon patches + all my USB keyboard patches
So I will add a CONFIG option to my "usb: kbd: implement special keys" patch to avoid the code increase.
With CONFIG_REGEX=n the image fits into the current board limit in all cases.
Soeren could you, please, evaluate if this configuration works with your board. It should only be relevant if multiple network interfaces are supported by U-Boot.
Best regards
Heinrich
PATCH v2 1/2 usb: composite: fix possible alignment issues
Quoted 54 lines · Heinrich Schuchardt
Quoted 41 lines · Marek Vasut
Quoted 40 lines · Heinrich Schuchardt
Quoted 28 lines · Simon Goldschmidt
Quoted 26 lines · Heinrich Schuchardt
Quoted 5 lines · Marek Vasut
Quoted 4 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Quoted 8 lines
Quoted 5 lines
Fix this by converting two functions to use unaligned access since packed
structures may be on an unaligned address, depending on USB hardware.Signed-off-by Simon Goldschmidt <[email protected]>Applied both, thanks.
With these two patches applied to origin/master GCC 9.2.1 does not produce warnings anymore but for
tbs2910_defconfig:u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytes
make: *** [Makefile:1159: u-boot.imx] Error 1
make: *** Deleting file 'u-boot.imx'Quoted 8 lines
Quoted 2 lines
So irrespective of my patches for the USB keyboard we need to address the size limit on TBS2910.
Is that due to my cleanup patches? Can you compare the size by compiling without them? That should work if you leave away the -Werror switch.
Regards,
SimonGCC 9.2.1 without your patches but with -Wno-address-of-packed-member:
Quoted 9 lines
Quoted 4 lines
u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytesI see, so you have additional options added to the build which trigger the size issue. It would be nice to mention that up front. Do you use any other extra options ?
Dear Marek,
Quoted 7 lines
Simon asked me to determine if origin/master exceeds the u-boot.imx size limit when compiled without his patches. The only way to do so is to suppress the build warnings.
-Wno-address-of-packed-member is the only option I added to origin/master. This option suppresses the build error that we get without Simon's patches.
For the record, with gcc 7.2, these types of fixes result in:
u-boot: add: 0/0, grow: 2/0 bytes: 36/0 (36)
function old new delta
collect_langs 76 100 +24
composite_setup 2440 2452 +12
spl-u-boot-spl: add: 0/0, grow: 2/0 bytes: 36/0 (36)
function old new delta
collect_langs 76 100 +24
composite_setup 2440 2452 +12
Signature · 8 lines
Tom -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 833 bytes Desc: not available URL: <http://lists.denx.de/pipermail/u-boot/attachments/20191122/cb0ea46b/attachment.sig>
PATCH v2 1/2 usb: composite: fix possible alignment issues
In reply to Heinrich SchuchardtQuoted 49 lines · Heinrich Schuchardt
Quoted 48 lines · Marek Vasut
Quoted 46 lines · Heinrich Schuchardt
Quoted 32 lines · Simon Goldschmidt
Quoted 27 lines · Heinrich Schuchardt
Quoted 5 lines · Marek Vasut
Quoted 4 lines · Simon Goldschmidt
Since upgrading to gcc9, warnings are issued:
"taking address of packed member of ‘...’ may result in an unaligned pointer value"Quoted 9 lines
Quoted 6 lines
Fix this by converting two functions to use unaligned access since packed
structures may be on an unaligned address, depending on USB hardware.Signed-off-by Simon Goldschmidt <[email protected]>Applied both, thanks.
With these two patches applied to origin/master GCC 9.2.1 does not produce warnings anymore but for
tbs2910_defconfig:u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytes
make: *** [Makefile:1159: u-boot.imx] Error 1
make: *** Deleting file 'u-boot.imx'Quoted 2 lines
So irrespective of my patches for the USB keyboard we need to address the size limit on TBS2910.
Quoted 6 lines
Is that due to my cleanup patches? Can you compare the size by compiling
without them? That should work if you leave away the -Werror switch.Regards,
SimonGCC 9.2.1 without your patches but with -Wno-address-of-packed-member:
u-boot.imx exceeds file size limit:
limit:0x5fc00bytes
actual:0x60c00bytes
excess:0x1000bytes
Quoted 6 lines
Quoted 4 lines
I see, so you have additional options added to the build which trigger the size issue. It would be nice to mention that up front. Do you use any other extra options ?
Dear Marek,
Hi,
Quoted 7 lines
Simon asked me to determine if origin/master exceeds the u-boot.imx size limit when compiled without his patches. The only way to do so is to suppress the build warnings.
-Wno-address-of-packed-member is the only option I added to origin/master. This option suppresses the build error that we get without Simon's patches.
But you somehow got this -Werror into the build too, right ?
Quoted 2 lines
The only other difference between Gitlab CI and my setup is that I use GCC 9.2.1 (arm-linux-gnueabihf-gcc on Debian Bullseye).
That's what I use too.
Quoted 10 lines
These are the sizes of u-boot.bin:
390080 bytes with Simon's patches
390080 bytes with Simon's patches + -Wno-address-of-packed-member 390064 bytes origin/master + -Wno-address-of-packed-member 386248 bytes with Simon's patches +CONFIG_REGEX=n390024 bytes with Simon patches +
"usb: kbd: simplify coding for arrow keys"
390440 bytes with Simon patches + all my USB keyboard patches
Quoted 2 lines
So I will add a CONFIG option to my "usb: kbd: implement special keys" patch to avoid the code increase.
There might be another option. How about improving the encoding of these escape sequences? There seem to be a lot of duplication, e.g. the leading '\e' is in all of them. So is the '[' . Maybe deduplicating those could help ?
I did a quick try by putting all the sequences into a table, then removed the \e and sent it explicitly with usb_kbd_put_queue(), and got ~100 bytes saved. And then each of those strings has trailing '\0', which I don't think is needed either, so that might be another few bytes saved.
$ arm-linux-gnueabi-readelf -s u-boot | sort -nk 3 | grep usb_kbd can help tracking down these bloat hotspots.
Quoted 6 lines
With
CONFIG_REGEX=nthe image fits into the current board limit in all cases.Soeren could you, please, evaluate if this configuration works with your board. It should only be relevant if multiple network interfaces are supported by U-Boot.
CONFIG_REGEX is used by setexpr to handle regexes on U-Boot command line. Hence, removing CONFIG_REGEX would degrade the board functionality and that is what I don't want to see.