Message ID | 20180918085053.lu2k46vqffzj37ah@verge.net.au (mailing list archive) |
---|---|
State | Not Applicable |
Delegated to: | Simon Horman |
Headers | show |
Series | [GIT,PULL,LTSI-4.14] LTSI-v4.14 Backport or I2C R-Car Fix | expand |
Hi Simon, On Tue, Sep 18, 2018 at 10:51 AM Simon Horman <horms@verge.net.au> wrote: > This is intended as a submission to LTSI-4.14. It is the backport > of a fixes for safe DMA buffer handling for the SH-Mobile I2C driver > and I2C code. All patches are present in v4.19-rc3. > > This pull-request is based on > "[GIT PULL LTSI-4.14] LTSI-v4.14 Backport or I2C R-Car Fix" > tagged as backport/v4.14.61/snapshot-to-v4.18+fixes-flattened, > which I have already sent a pull-request for. > > There are 10 patches. > > I have performed build testing of this backports on a wide range of > defconfigs and I am not aware of any regressions over v4.14.40 (the > baseline chosen when this work began). > > > The following changes since commit 4d4605e5c137ed9a53582e573118cbc16b82cbf1: > > i2c: rcar: implement STOP and REP_START according to docs (2018-08-28 13:35:06 +0200) > > are available in the git repository at: > > https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas-backport.git backport/v4.14.61/snapshot-to-v4.18+fixes-v2-flattened > > for you to fetch changes up to 720043a4efbb3356db116d084c7b877ad41ee51a: > > i2c: sh_mobile: fix leak when using DMA bounce buffer (2018-09-17 15:54:30 +0200) Thank, this all looks fine to me. I subjected this to the same testing I do for each renesas-drivers release. I have detected no regressions[*]. [*] The only regression I'm aware of is a regression in 4.14-stable, which can be fixed by "tick/nohz: Prevent bogus softirq pending warning". (https://lore.kernel.org/patchwork/patch/979451/). Gr{oetje,eeting}s, Geert
On Tue, Sep 18, 2018 at 12:07:50PM +0200, Geert Uytterhoeven wrote: > Hi Simon, > > On Tue, Sep 18, 2018 at 10:51 AM Simon Horman <horms@verge.net.au> wrote: > > This is intended as a submission to LTSI-4.14. It is the backport > > of a fixes for safe DMA buffer handling for the SH-Mobile I2C driver > > and I2C code. All patches are present in v4.19-rc3. > > > > This pull-request is based on > > "[GIT PULL LTSI-4.14] LTSI-v4.14 Backport or I2C R-Car Fix" > > tagged as backport/v4.14.61/snapshot-to-v4.18+fixes-flattened, > > which I have already sent a pull-request for. > > > > There are 10 patches. > > > > I have performed build testing of this backports on a wide range of > > defconfigs and I am not aware of any regressions over v4.14.40 (the > > baseline chosen when this work began). > > > > > > The following changes since commit 4d4605e5c137ed9a53582e573118cbc16b82cbf1: > > > > i2c: rcar: implement STOP and REP_START according to docs (2018-08-28 13:35:06 +0200) > > > > are available in the git repository at: > > > > https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas-backport.git backport/v4.14.61/snapshot-to-v4.18+fixes-v2-flattened > > > > for you to fetch changes up to 720043a4efbb3356db116d084c7b877ad41ee51a: > > > > i2c: sh_mobile: fix leak when using DMA bounce buffer (2018-09-17 15:54:30 +0200) > > Thank, this all looks fine to me. > > I subjected this to the same testing I do for each renesas-drivers release. > I have detected no regressions[*]. > > [*] The only regression I'm aware of is a regression in 4.14-stable, which can > be fixed by "tick/nohz: Prevent bogus softirq pending warning". > (https://lore.kernel.org/patchwork/patch/979451/). Is that something we want to add to LTSI-4.14?
Hi Simon, On Wed, Sep 19, 2018 at 11:08 AM Simon Horman <horms@verge.net.au> wrote: > On Tue, Sep 18, 2018 at 12:07:50PM +0200, Geert Uytterhoeven wrote: > > [*] The only regression I'm aware of is a regression in 4.14-stable, which can > > be fixed by "tick/nohz: Prevent bogus softirq pending warning". > > (https://lore.kernel.org/patchwork/patch/979451/). > > Is that something we want to add to LTSI-4.14? As that is a fix for a regression introduced in v4.14.62, I think it should be fixed in 4.14-stable instead (and hopefully be fixed automatically in v4.14.x-ltsi rc2 or final). Gr{oetje,eeting}s, Geert
On Wed, Sep 19, 2018 at 11:13:14AM +0200, Geert Uytterhoeven wrote: > Hi Simon, > > On Wed, Sep 19, 2018 at 11:08 AM Simon Horman <horms@verge.net.au> wrote: > > On Tue, Sep 18, 2018 at 12:07:50PM +0200, Geert Uytterhoeven wrote: > > > [*] The only regression I'm aware of is a regression in 4.14-stable, which can > > > be fixed by "tick/nohz: Prevent bogus softirq pending warning". > > > (https://lore.kernel.org/patchwork/patch/979451/). > > > > Is that something we want to add to LTSI-4.14? > > As that is a fix for a regression introduced in v4.14.62, I think it should > be fixed in 4.14-stable instead (and hopefully be fixed automatically in > v4.14.x-ltsi rc2 or final). Ok, that sounds reasonable to me. Let me know if you want to revisit this at some point.
Hi Simon, On Tue, Sep 18, 2018 at 10:51 AM Simon Horman <horms@verge.net.au> wrote: > This is intended as a submission to LTSI-4.14. It is the backport > of a fixes for safe DMA buffer handling for the SH-Mobile I2C driver > and I2C code. All patches are present in v4.19-rc3. > > This pull-request is based on > "[GIT PULL LTSI-4.14] LTSI-v4.14 Backport or I2C R-Car Fix" > tagged as backport/v4.14.61/snapshot-to-v4.18+fixes-flattened, > which I have already sent a pull-request for. > > There are 10 patches. > > I have performed build testing of this backports on a wide range of > defconfigs and I am not aware of any regressions over v4.14.40 (the > baseline chosen when this work began). > > > The following changes since commit 4d4605e5c137ed9a53582e573118cbc16b82cbf1: > > i2c: rcar: implement STOP and REP_START according to docs (2018-08-28 13:35:06 +0200) > > are available in the git repository at: > > https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas-backport.git backport/v4.14.61/snapshot-to-v4.18+fixes-v2-flattened > > for you to fetch changes up to 720043a4efbb3356db116d084c7b877ad41ee51a: > > i2c: sh_mobile: fix leak when using DMA bounce buffer (2018-09-17 15:54:30 +0200) > > ---------------------------------------------------------------- > Second Round of LTSI-v4.14 Backports of I2C R-Car Fixes > > Base: > * v4.14.61 > * Backports of components for Renesas SoCs to v4.18 > * First round of Backports of I2C R-Car Fixes > > Backport of post-v4.18 fix for i2c-shmobile and I2C core > The focus of these fixes is to make DMA buffer handling safe > > ---------------------------------------------------------------- > Peter Rosin (1): > i2c: smbus: kill memory leak on emulated and failed DMA SMBus xfers > > Wenwen Wang (1): > i2c: core: smbus: fix a potential missing-check bug > > Wolfram Sang (8): > i2c: dev: mark RDWR buffers as DMA_SAFE > i2c: refactor i2c_master_{send_recv} > i2c: add i2c_master_{send|recv}_dmasafe > i2c: smbus: use DMA safe buffers for emulated SMBus transactions > i2c: add docs to clarify DMA handling > i2c: refactor function to release a DMA safe buffer > i2c: sh_mobile: define start_ch() void as it only returns 0 anyhow Somehow I missed that the above introduces a (rightful) compiler warning: drivers/i2c/busses/i2c-sh_mobile.c: In function ‘start_ch’: drivers/i2c/busses/i2c-sh_mobile.c:616:10: warning: ‘return’ with a value, in function returning void return -EOPNOTSUPP; ^ drivers/i2c/busses/i2c-sh_mobile.c:611:13: note: declared here static void start_ch(struct sh_mobile_i2c_data *pd, struct i2c_msg *usr_msg, ^~~~~~~~ as it depends on commit 7957c2e276ee73e6 ("i2c: sh_mobile: use core to detect 'no zero length read' quirk"), which was not backported. I think the easiest solution is to just drop "i2c: sh_mobile: define start_ch() void as it only returns 0 anyhow", as it was meant not to introduce any behavioral changes, and nothing depends on it. Do you agree? Thanks! > i2c: sh_mobile: fix leak when using DMA bounce buffer Gr{oetje,eeting}s, Geert
On Mon, Oct 01, 2018 at 12:08:38PM +0200, Geert Uytterhoeven wrote: > Hi Simon, > > On Tue, Sep 18, 2018 at 10:51 AM Simon Horman <horms@verge.net.au> wrote: > > This is intended as a submission to LTSI-4.14. It is the backport > > of a fixes for safe DMA buffer handling for the SH-Mobile I2C driver > > and I2C code. All patches are present in v4.19-rc3. > > > > This pull-request is based on > > "[GIT PULL LTSI-4.14] LTSI-v4.14 Backport or I2C R-Car Fix" > > tagged as backport/v4.14.61/snapshot-to-v4.18+fixes-flattened, > > which I have already sent a pull-request for. > > > > There are 10 patches. > > > > I have performed build testing of this backports on a wide range of > > defconfigs and I am not aware of any regressions over v4.14.40 (the > > baseline chosen when this work began). > > > > > > The following changes since commit 4d4605e5c137ed9a53582e573118cbc16b82cbf1: > > > > i2c: rcar: implement STOP and REP_START according to docs (2018-08-28 13:35:06 +0200) > > > > are available in the git repository at: > > > > https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas-backport.git backport/v4.14.61/snapshot-to-v4.18+fixes-v2-flattened > > > > for you to fetch changes up to 720043a4efbb3356db116d084c7b877ad41ee51a: > > > > i2c: sh_mobile: fix leak when using DMA bounce buffer (2018-09-17 15:54:30 +0200) > > > > ---------------------------------------------------------------- > > Second Round of LTSI-v4.14 Backports of I2C R-Car Fixes > > > > Base: > > * v4.14.61 > > * Backports of components for Renesas SoCs to v4.18 > > * First round of Backports of I2C R-Car Fixes > > > > Backport of post-v4.18 fix for i2c-shmobile and I2C core > > The focus of these fixes is to make DMA buffer handling safe > > > > ---------------------------------------------------------------- > > Peter Rosin (1): > > i2c: smbus: kill memory leak on emulated and failed DMA SMBus xfers > > > > Wenwen Wang (1): > > i2c: core: smbus: fix a potential missing-check bug > > > > Wolfram Sang (8): > > i2c: dev: mark RDWR buffers as DMA_SAFE > > i2c: refactor i2c_master_{send_recv} > > i2c: add i2c_master_{send|recv}_dmasafe > > i2c: smbus: use DMA safe buffers for emulated SMBus transactions > > i2c: add docs to clarify DMA handling > > i2c: refactor function to release a DMA safe buffer > > i2c: sh_mobile: define start_ch() void as it only returns 0 anyhow > > Somehow I missed that the above introduces a (rightful) compiler warning: > > drivers/i2c/busses/i2c-sh_mobile.c: In function ‘start_ch’: > drivers/i2c/busses/i2c-sh_mobile.c:616:10: warning: ‘return’ with > a value, in function returning void > return -EOPNOTSUPP; > ^ > drivers/i2c/busses/i2c-sh_mobile.c:611:13: note: declared here > static void start_ch(struct sh_mobile_i2c_data *pd, struct > i2c_msg *usr_msg, > ^~~~~~~~ > > as it depends on commit 7957c2e276ee73e6 ("i2c: sh_mobile: use core to > detect 'no zero length read' quirk"), which was not backported. > > I think the easiest solution is to just drop "i2c: sh_mobile: define > start_ch() void as it only returns 0 anyhow", as it was meant not to > introduce any behavioral changes, and nothing depends on it. > > Do you agree? > Thanks! Likewise, thanks for pointing this out. Your proposed solution looks good to me. Have you tested it? > > > i2c: sh_mobile: fix leak when using DMA bounce buffer > > Gr{oetje,eeting}s, > > Geert > > -- > Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org > > In personal conversations with technical people, I call myself a hacker. But > when I'm talking to journalists I just say "programmer" or something like that. > -- Linus Torvalds >
Hi Simon, On Mon, Oct 1, 2018 at 3:52 PM Simon Horman <horms@verge.net.au> wrote: > On Mon, Oct 01, 2018 at 12:08:38PM +0200, Geert Uytterhoeven wrote: > > On Tue, Sep 18, 2018 at 10:51 AM Simon Horman <horms@verge.net.au> wrote: > > > This is intended as a submission to LTSI-4.14. It is the backport > > > of a fixes for safe DMA buffer handling for the SH-Mobile I2C driver > > > and I2C code. All patches are present in v4.19-rc3. > > > > > > This pull-request is based on > > > "[GIT PULL LTSI-4.14] LTSI-v4.14 Backport or I2C R-Car Fix" > > > tagged as backport/v4.14.61/snapshot-to-v4.18+fixes-flattened, > > > which I have already sent a pull-request for. > > > > > > There are 10 patches. > > > > > > I have performed build testing of this backports on a wide range of > > > defconfigs and I am not aware of any regressions over v4.14.40 (the > > > baseline chosen when this work began). > > > > > > The following changes since commit 4d4605e5c137ed9a53582e573118cbc16b82cbf1: > > > > > > i2c: rcar: implement STOP and REP_START according to docs (2018-08-28 13:35:06 +0200) > > > > > > are available in the git repository at: > > > > > > https://git.kernel.org/pub/scm/linux/kernel/git/horms/renesas-backport.git backport/v4.14.61/snapshot-to-v4.18+fixes-v2-flattened > > > > > > for you to fetch changes up to 720043a4efbb3356db116d084c7b877ad41ee51a: > > > > > > i2c: sh_mobile: fix leak when using DMA bounce buffer (2018-09-17 15:54:30 +0200) > > > > > > ---------------------------------------------------------------- > > > Second Round of LTSI-v4.14 Backports of I2C R-Car Fixes > > > > > > Base: > > > * v4.14.61 > > > * Backports of components for Renesas SoCs to v4.18 > > > * First round of Backports of I2C R-Car Fixes > > > > > > Backport of post-v4.18 fix for i2c-shmobile and I2C core > > > The focus of these fixes is to make DMA buffer handling safe > > > > > > ---------------------------------------------------------------- > > > Peter Rosin (1): > > > i2c: smbus: kill memory leak on emulated and failed DMA SMBus xfers > > > > > > Wenwen Wang (1): > > > i2c: core: smbus: fix a potential missing-check bug > > > > > > Wolfram Sang (8): > > > i2c: dev: mark RDWR buffers as DMA_SAFE > > > i2c: refactor i2c_master_{send_recv} > > > i2c: add i2c_master_{send|recv}_dmasafe > > > i2c: smbus: use DMA safe buffers for emulated SMBus transactions > > > i2c: add docs to clarify DMA handling > > > i2c: refactor function to release a DMA safe buffer > > > i2c: sh_mobile: define start_ch() void as it only returns 0 anyhow > > > > Somehow I missed that the above introduces a (rightful) compiler warning: > > > > drivers/i2c/busses/i2c-sh_mobile.c: In function ‘start_ch’: > > drivers/i2c/busses/i2c-sh_mobile.c:616:10: warning: ‘return’ with > > a value, in function returning void > > return -EOPNOTSUPP; > > ^ > > drivers/i2c/busses/i2c-sh_mobile.c:611:13: note: declared here > > static void start_ch(struct sh_mobile_i2c_data *pd, struct > > i2c_msg *usr_msg, > > ^~~~~~~~ > > > > as it depends on commit 7957c2e276ee73e6 ("i2c: sh_mobile: use core to > > detect 'no zero length read' quirk"), which was not backported. > > > > I think the easiest solution is to just drop "i2c: sh_mobile: define > > start_ch() void as it only returns 0 anyhow", as it was meant not to > > introduce any behavioral changes, and nothing depends on it. > > > > Do you agree? > > Thanks! > > Likewise, thanks for pointing this out. > > Your proposed solution looks good to me. > Have you tested it? More or less, the error case is not easily triggered, and the code should behave as before. Gr{oetje,eeting}s, Geert