| Message ID | 20190313111545.1812261-1-hs@denx.de |
|---|---|
| State | Accepted |
| Commit | 9ba5e5bc261a16f51662490da0cf620dc7f29013 |
| Delegated to: | Tom Rini |
| Headers | show |
| Series | [U-Boot] Revert "env: add spi_flash_read_env function" | expand |
On 13/03/19 12:15, Heiko Schocher wrote: > This reverts commit 9a9d66f5eff0f443de4c2c6ca3e27771ed14b1b4. > > because it breaks fw_setenv and U-Boot interworking, if > U-Boot environment is stored in a SPI-NOR. > > Reproduce it with: > boot linux with empty Environment and store a variable > with fw_setenv into it, the Environment is now filled > with 0xff: > > root@ckey5e:10:8e:~# hexdump -C /dev/mtd4 > 00000000 e9 e8 07 fa 01 62 6f 6f 74 63 6d 64 3d 72 75 6e |.....bootcmd=run| > [...] > 00000f30 7d 00 75 62 69 62 6f 6f 74 76 6f 6c 3d 32 00 00 |}.ubibootvol=2..| > 00000f40 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................| > > Boot now U-Boot prints: > > Loading Environment from SPI Flash... SF: Detected s25fl128l with page size 256 Bytes, erase size 4 KiB, total 16 MiB > *** Warning - bad CRC, using default environment > > Reason is the above commit, as it only reads until \0\0 > is found, and assumes the rest of the Environment > space is filled with 0x00, which is not the case when > saving an Environment under linux with fw_setenv. > > Signed-off-by: Heiko Schocher <hs@denx.de> > --- Acked-by: Stefano Babic <sbabic@denx.de> In fact, the commit breaks the API between U-Boot and Linux and optimizes a specific use case in U-Boot only. The API foresees that there is an environment of a specific size without bothering (in linux) where it is stored. It is strictly required that the environment has the same format for all underlying storage - it must be possible to simply copy the environment. Commit breaks tools like mkenvimage, too. Best regards, Stefano Babic > I already posted a fix: > https://lists.denx.de/pipermail/u-boot/2019-January/355148.html > which I withdrawed: > https://lists.denx.de/pipermail/u-boot/2019-January/355529.html > > because I did not recognised the correlation between fw_setenv > in linux context and saveenv in U-Boot context. > > I think now, best is to revert this commit. > > Comments? > > This patch drops checkpatch warnings: > env/sf.c:120: check: Alignment should match open parenthesis > env/sf.c:224: check: Alignment should match open parenthesis > env/sf.c:281: check: Alignment should match open parenthesis > > but as it is a revert I did not correct them. > > env/sf.c | 56 +++++++++++--------------------------------------------- > 1 file changed, 11 insertions(+), 45 deletions(-) > > diff --git a/env/sf.c b/env/sf.c > index b3dec82c35..23cbad5d88 100644 > --- a/env/sf.c > +++ b/env/sf.c > @@ -81,40 +81,6 @@ static int setup_flash_device(void) > return 0; > } > > -static int is_end(const char *addr, size_t size) > -{ > - /* The end of env variables is marked by '\0\0' */ > - int i = 0; > - > - for (i = 0; i < size - 1; ++i) > - if (addr[i] == 0x0 && addr[i + 1] == 0x0) > - return 1; > - return 0; > -} > - > -static int spi_flash_read_env(struct spi_flash *flash, u32 offset, size_t len, > - void *buf) > -{ > - u32 addr = 0; > - u32 page_size = flash->page_size; > - > - memset(buf, 0x0, len); > - for (int i = 0; i < len / page_size; ++i) { > - int ret = spi_flash_read(flash, offset, page_size, > - &((char *)buf)[addr]); > - > - if (ret < 0) > - return ret; > - > - if (is_end(&((char *)buf)[addr], page_size)) > - return 0; > - > - addr += page_size; > - offset += page_size; > - } > - return 0; > -} > - > #if defined(CONFIG_ENV_OFFSET_REDUND) > #ifdef CMD_SAVEENV > static int env_sf_save(void) > @@ -150,8 +116,8 @@ static int env_sf_save(void) > ret = -ENOMEM; > goto done; > } > - ret = spi_flash_read_env(env_flash, saved_offset, > - saved_size, saved_buffer); > + ret = spi_flash_read(env_flash, saved_offset, > + saved_size, saved_buffer); > if (ret) > goto done; > } > @@ -217,10 +183,10 @@ static int env_sf_load(void) > if (ret) > goto out; > > - read1_fail = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET, > - CONFIG_ENV_SIZE, tmp_env1); > - read2_fail = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET_REDUND, > - CONFIG_ENV_SIZE, tmp_env2); > + read1_fail = spi_flash_read(env_flash, CONFIG_ENV_OFFSET, > + CONFIG_ENV_SIZE, tmp_env1); > + read2_fail = spi_flash_read(env_flash, CONFIG_ENV_OFFSET_REDUND, > + CONFIG_ENV_SIZE, tmp_env2); > > ret = env_import_redund((char *)tmp_env1, read1_fail, (char *)tmp_env2, > read2_fail); > @@ -254,8 +220,8 @@ static int env_sf_save(void) > if (!saved_buffer) > goto done; > > - ret = spi_flash_read_env(env_flash, saved_offset, > - saved_size, saved_buffer); > + ret = spi_flash_read(env_flash, saved_offset, > + saved_size, saved_buffer); > if (ret) > goto done; > } > @@ -311,10 +277,10 @@ static int env_sf_load(void) > if (ret) > goto out; > > - ret = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET, CONFIG_ENV_SIZE, > - buf); > + ret = spi_flash_read(env_flash, > + CONFIG_ENV_OFFSET, CONFIG_ENV_SIZE, buf); > if (ret) { > - set_default_env("spi_flash_read_env() failed", 0); > + set_default_env("spi_flash_read() failed", 0); > goto err_read; > } > >
Hi Heiko, I managed to reproduce the issue that you described. Don't you think it is a little bit too harsh to remove the commit completely? I am not sure but how many cases are where UBoot doesn't store anything in flash? And then lets the linux to update the flash. Or is wouldn't better to update fw_setenv to set the entire flash to 0x0 when it detects that the CRC error? The same way how the saveenv it is doing in UBoot? /Horatiu The 03/13/2019 12:15, Heiko Schocher wrote: > This reverts commit 9a9d66f5eff0f443de4c2c6ca3e27771ed14b1b4. > > because it breaks fw_setenv and U-Boot interworking, if > U-Boot environment is stored in a SPI-NOR. > > Reproduce it with: > boot linux with empty Environment and store a variable > with fw_setenv into it, the Environment is now filled > with 0xff: > > root@ckey5e:10:8e:~# hexdump -C /dev/mtd4 > 00000000 e9 e8 07 fa 01 62 6f 6f 74 63 6d 64 3d 72 75 6e |.....bootcmd=run| > [...] > 00000f30 7d 00 75 62 69 62 6f 6f 74 76 6f 6c 3d 32 00 00 |}.ubibootvol=2..| > 00000f40 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................| > > Boot now U-Boot prints: > > Loading Environment from SPI Flash... SF: Detected s25fl128l with page size 256 Bytes, erase size 4 KiB, total 16 MiB > *** Warning - bad CRC, using default environment > > Reason is the above commit, as it only reads until \0\0 > is found, and assumes the rest of the Environment > space is filled with 0x00, which is not the case when > saving an Environment under linux with fw_setenv. > > Signed-off-by: Heiko Schocher <hs@denx.de> > --- > I already posted a fix: > https://lists.denx.de/pipermail/u-boot/2019-January/355148.html > which I withdrawed: > https://lists.denx.de/pipermail/u-boot/2019-January/355529.html > > because I did not recognised the correlation between fw_setenv > in linux context and saveenv in U-Boot context. > > I think now, best is to revert this commit. > > Comments? > > This patch drops checkpatch warnings: > env/sf.c:120: check: Alignment should match open parenthesis > env/sf.c:224: check: Alignment should match open parenthesis > env/sf.c:281: check: Alignment should match open parenthesis > > but as it is a revert I did not correct them. > > env/sf.c | 56 +++++++++++--------------------------------------------- > 1 file changed, 11 insertions(+), 45 deletions(-) > > diff --git a/env/sf.c b/env/sf.c > index b3dec82c35..23cbad5d88 100644 > --- a/env/sf.c > +++ b/env/sf.c > @@ -81,40 +81,6 @@ static int setup_flash_device(void) > return 0; > } > > -static int is_end(const char *addr, size_t size) > -{ > - /* The end of env variables is marked by '\0\0' */ > - int i = 0; > - > - for (i = 0; i < size - 1; ++i) > - if (addr[i] == 0x0 && addr[i + 1] == 0x0) > - return 1; > - return 0; > -} > - > -static int spi_flash_read_env(struct spi_flash *flash, u32 offset, size_t len, > - void *buf) > -{ > - u32 addr = 0; > - u32 page_size = flash->page_size; > - > - memset(buf, 0x0, len); > - for (int i = 0; i < len / page_size; ++i) { > - int ret = spi_flash_read(flash, offset, page_size, > - &((char *)buf)[addr]); > - > - if (ret < 0) > - return ret; > - > - if (is_end(&((char *)buf)[addr], page_size)) > - return 0; > - > - addr += page_size; > - offset += page_size; > - } > - return 0; > -} > - > #if defined(CONFIG_ENV_OFFSET_REDUND) > #ifdef CMD_SAVEENV > static int env_sf_save(void) > @@ -150,8 +116,8 @@ static int env_sf_save(void) > ret = -ENOMEM; > goto done; > } > - ret = spi_flash_read_env(env_flash, saved_offset, > - saved_size, saved_buffer); > + ret = spi_flash_read(env_flash, saved_offset, > + saved_size, saved_buffer); > if (ret) > goto done; > } > @@ -217,10 +183,10 @@ static int env_sf_load(void) > if (ret) > goto out; > > - read1_fail = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET, > - CONFIG_ENV_SIZE, tmp_env1); > - read2_fail = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET_REDUND, > - CONFIG_ENV_SIZE, tmp_env2); > + read1_fail = spi_flash_read(env_flash, CONFIG_ENV_OFFSET, > + CONFIG_ENV_SIZE, tmp_env1); > + read2_fail = spi_flash_read(env_flash, CONFIG_ENV_OFFSET_REDUND, > + CONFIG_ENV_SIZE, tmp_env2); > > ret = env_import_redund((char *)tmp_env1, read1_fail, (char *)tmp_env2, > read2_fail); > @@ -254,8 +220,8 @@ static int env_sf_save(void) > if (!saved_buffer) > goto done; > > - ret = spi_flash_read_env(env_flash, saved_offset, > - saved_size, saved_buffer); > + ret = spi_flash_read(env_flash, saved_offset, > + saved_size, saved_buffer); > if (ret) > goto done; > } > @@ -311,10 +277,10 @@ static int env_sf_load(void) > if (ret) > goto out; > > - ret = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET, CONFIG_ENV_SIZE, > - buf); > + ret = spi_flash_read(env_flash, > + CONFIG_ENV_OFFSET, CONFIG_ENV_SIZE, buf); > if (ret) { > - set_default_env("spi_flash_read_env() failed", 0); > + set_default_env("spi_flash_read() failed", 0); > goto err_read; > } > > -- > 2.17.2 >
On Wed, Mar 13, 2019 at 12:15:45PM +0100, Heiko Schocher wrote: > This reverts commit 9a9d66f5eff0f443de4c2c6ca3e27771ed14b1b4. > > because it breaks fw_setenv and U-Boot interworking, if > U-Boot environment is stored in a SPI-NOR. > > Reproduce it with: > boot linux with empty Environment and store a variable > with fw_setenv into it, the Environment is now filled > with 0xff: > > root@ckey5e:10:8e:~# hexdump -C /dev/mtd4 > 00000000 e9 e8 07 fa 01 62 6f 6f 74 63 6d 64 3d 72 75 6e |.....bootcmd=run| > [...] > 00000f30 7d 00 75 62 69 62 6f 6f 74 76 6f 6c 3d 32 00 00 |}.ubibootvol=2..| > 00000f40 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................| > > Boot now U-Boot prints: > > Loading Environment from SPI Flash... SF: Detected s25fl128l with page size 256 Bytes, erase size 4 KiB, total 16 MiB > *** Warning - bad CRC, using default environment > > Reason is the above commit, as it only reads until \0\0 > is found, and assumes the rest of the Environment > space is filled with 0x00, which is not the case when > saving an Environment under linux with fw_setenv. > > Signed-off-by: Heiko Schocher <hs@denx.de> > Acked-by: Stefano Babic <sbabic@denx.de> Applied to u-boot/master, thanks!
Hello Horatiu, Am 14.03.2019 um 13:53 schrieb Horatiu Vultur: > Hi Heiko, > > I managed to reproduce the issue that you described. Fine, thanks for testing! > Don't you think it is a little bit too harsh to remove the commit > completely? I am unsure here ... > I am not sure but how many cases are where UBoot doesn't store anything > in flash? And then lets the linux to update the flash. I think that can happen very often. In my case for example we boot SPL/U-Boot with usb loader, than an initsystem with kernel/dtb/ramdisk and start swupdate for installing all sources (including setup the Environment with fw_setenv)... > Or is wouldn't better to update fw_setenv to set the entire flash to > 0x0 when it detects that the CRC error? The same way how the saveenv it > is doing in UBoot? Indeed, just reproduced this on my board. Saving the Environment with saveenv fills with 0x0 and saving with fw_setenv fills with 0xff. This is no problem, until your optimization, and I think, we cannot change this here without breaking a lot of boards, running fine. I am here also on Stefanos side always to read the hole Environment space, because we cannot be sure, with what the empty space is filled up. Back to your intention for this patch, stated in the commit message: """ This is an optimization for large environments that contain few bytes environment variables. In this case it doesn't need to read the entire environment and only few pages. """ Hmm... why you use a big Environment, if you only have "few bytes" of variables ? Simply set CONFIG_ENV_SIZE to a smaller value than your CONFIG_ENV_SECT_SIZE Is this not possible for you ? bye, Heiko
Hi Heiko, Sorry, for late reply. I have seen that this patch was accepted, so I have only few things bellow. The 03/15/2019 05:34, Heiko Schocher wrote: > External E-Mail > > > Hello Horatiu, > > Am 14.03.2019 um 13:53 schrieb Horatiu Vultur: > > Hi Heiko, > > > > I managed to reproduce the issue that you described. > > Fine, thanks for testing! > > > Don't you think it is a little bit too harsh to remove the commit > > completely? > > I am unsure here ... > > > I am not sure but how many cases are where UBoot doesn't store anything > > in flash? And then lets the linux to update the flash. > > I think that can happen very often. In my case for example we boot > SPL/U-Boot with usb loader, than an initsystem with kernel/dtb/ramdisk > and start swupdate for installing all sources (including setup > the Environment with fw_setenv)... > > > Or is wouldn't better to update fw_setenv to set the entire flash to > > 0x0 when it detects that the CRC error? The same way how the saveenv it > > is doing in UBoot? > > Indeed, just reproduced this on my board. Saving the Environment > with saveenv fills with 0x0 and saving with fw_setenv fills with > 0xff. > > This is no problem, until your optimization, and I think, we cannot > change this here without breaking a lot of boards, running fine. > > I am here also on Stefanos side always to read the hole Environment > space, because we cannot be sure, with what the empty space is filled > up. > > Back to your intention for this patch, stated in the commit message: > """ > This is an optimization for large environments that contain few bytes > environment variables. In this case it doesn't need to read the entire > environment and only few pages. > """ > > Hmm... why you use a big Environment, if you only have "few bytes" of > variables ? We choose a big Environment because different boards can have different flashes which can have different page size. Therefore we chose for the environmnet size to be the least common multiple. > > Simply set CONFIG_ENV_SIZE to a smaller value than your CONFIG_ENV_SECT_SIZE > > Is this not possible for you ? Thanks for the suggestion. I have done few tests with this option and improves the speed setting the CONFIG_ENV_SIZE to something smaller than CONFIG_ENV_SECT_SIZE. > > bye, > Heiko > -- > DENX Software Engineering GmbH, Managing Director: Wolfgang Denk > HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany > Phone: +49-8142-66989-52 Fax: +49-8142-66989-80 Email: hs@denx.de
diff --git a/env/sf.c b/env/sf.c index b3dec82c35..23cbad5d88 100644 --- a/env/sf.c +++ b/env/sf.c @@ -81,40 +81,6 @@ static int setup_flash_device(void) return 0; } -static int is_end(const char *addr, size_t size) -{ - /* The end of env variables is marked by '\0\0' */ - int i = 0; - - for (i = 0; i < size - 1; ++i) - if (addr[i] == 0x0 && addr[i + 1] == 0x0) - return 1; - return 0; -} - -static int spi_flash_read_env(struct spi_flash *flash, u32 offset, size_t len, - void *buf) -{ - u32 addr = 0; - u32 page_size = flash->page_size; - - memset(buf, 0x0, len); - for (int i = 0; i < len / page_size; ++i) { - int ret = spi_flash_read(flash, offset, page_size, - &((char *)buf)[addr]); - - if (ret < 0) - return ret; - - if (is_end(&((char *)buf)[addr], page_size)) - return 0; - - addr += page_size; - offset += page_size; - } - return 0; -} - #if defined(CONFIG_ENV_OFFSET_REDUND) #ifdef CMD_SAVEENV static int env_sf_save(void) @@ -150,8 +116,8 @@ static int env_sf_save(void) ret = -ENOMEM; goto done; } - ret = spi_flash_read_env(env_flash, saved_offset, - saved_size, saved_buffer); + ret = spi_flash_read(env_flash, saved_offset, + saved_size, saved_buffer); if (ret) goto done; } @@ -217,10 +183,10 @@ static int env_sf_load(void) if (ret) goto out; - read1_fail = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET, - CONFIG_ENV_SIZE, tmp_env1); - read2_fail = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET_REDUND, - CONFIG_ENV_SIZE, tmp_env2); + read1_fail = spi_flash_read(env_flash, CONFIG_ENV_OFFSET, + CONFIG_ENV_SIZE, tmp_env1); + read2_fail = spi_flash_read(env_flash, CONFIG_ENV_OFFSET_REDUND, + CONFIG_ENV_SIZE, tmp_env2); ret = env_import_redund((char *)tmp_env1, read1_fail, (char *)tmp_env2, read2_fail); @@ -254,8 +220,8 @@ static int env_sf_save(void) if (!saved_buffer) goto done; - ret = spi_flash_read_env(env_flash, saved_offset, - saved_size, saved_buffer); + ret = spi_flash_read(env_flash, saved_offset, + saved_size, saved_buffer); if (ret) goto done; } @@ -311,10 +277,10 @@ static int env_sf_load(void) if (ret) goto out; - ret = spi_flash_read_env(env_flash, CONFIG_ENV_OFFSET, CONFIG_ENV_SIZE, - buf); + ret = spi_flash_read(env_flash, + CONFIG_ENV_OFFSET, CONFIG_ENV_SIZE, buf); if (ret) { - set_default_env("spi_flash_read_env() failed", 0); + set_default_env("spi_flash_read() failed", 0); goto err_read; }
This reverts commit 9a9d66f5eff0f443de4c2c6ca3e27771ed14b1b4. because it breaks fw_setenv and U-Boot interworking, if U-Boot environment is stored in a SPI-NOR. Reproduce it with: boot linux with empty Environment and store a variable with fw_setenv into it, the Environment is now filled with 0xff: root@ckey5e:10:8e:~# hexdump -C /dev/mtd4 00000000 e9 e8 07 fa 01 62 6f 6f 74 63 6d 64 3d 72 75 6e |.....bootcmd=run| [...] 00000f30 7d 00 75 62 69 62 6f 6f 74 76 6f 6c 3d 32 00 00 |}.ubibootvol=2..| 00000f40 00 ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff |................| Boot now U-Boot prints: Loading Environment from SPI Flash... SF: Detected s25fl128l with page size 256 Bytes, erase size 4 KiB, total 16 MiB *** Warning - bad CRC, using default environment Reason is the above commit, as it only reads until \0\0 is found, and assumes the rest of the Environment space is filled with 0x00, which is not the case when saving an Environment under linux with fw_setenv. Signed-off-by: Heiko Schocher <hs@denx.de> --- I already posted a fix: https://lists.denx.de/pipermail/u-boot/2019-January/355148.html which I withdrawed: https://lists.denx.de/pipermail/u-boot/2019-January/355529.html because I did not recognised the correlation between fw_setenv in linux context and saveenv in U-Boot context. I think now, best is to revert this commit. Comments? This patch drops checkpatch warnings: env/sf.c:120: check: Alignment should match open parenthesis env/sf.c:224: check: Alignment should match open parenthesis env/sf.c:281: check: Alignment should match open parenthesis but as it is a revert I did not correct them. env/sf.c | 56 +++++++++++--------------------------------------------- 1 file changed, 11 insertions(+), 45 deletions(-)