Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
cc: stable ab2a4bf83902c170d29ba130a8abb5f9d90559e1 USB: don't free bandwidth_mutex too early c538b9436751a0be2e1246b48353bc23156bdbcc pinctrl: intel: Only restore pins that are used by the driver
Seem useful; similar patches already in stable 3a020a723c65eb8ffa7c237faca26521a024e582 crypto: ghash-clmulni - Fix load failure 1a07834024dfca5c4bed5de8f8714306e0a11836 crypto: cryptd - Assign statesize properly ddef482420b1ba8ec45e6123a7e8d3f67b21e5e3 crypto: mcryptd - Fix load failure
Seem useful: 83430833b4d4a9c9b23964babbeb1f36450f8136 cxlflash: Increase cmd_per_lun for better throughput e34fbbac669de0b7fb7803929d0477f35f6e2833 ACPI / video: skip evaluating _DOD when it does not exist bcb48cca23ec9852739e4a464307fa29515bbe48 pinctrl: cherryview: Do not mask all interrupts in probe 77c0c9735bc0ba5898e637a3a20d6bcb50e3f67d Drivers: hv: balloon: don't crash when memory is added in non-sorted order a9f61ca793becabdefab03b77568d6c6f8c1bc79 Drivers: hv: avoid vfree() on crash 707e59ba494372a90d245f18b0c78982caa88e48 xen/qspinlock: Don't kick CPU if IRQ is not initialized 708e75a3ee750dce1072134e630d66c4e6eaf63c KVM: PPC: Book3S PR: Fix illegal opcode emulation dba599091c191d209b1499511a524ad9657c0e5a s390/pci: fix use after free in dma_init
A few more 'maybe's' that aren't marked cc: stable, but seemed useful; I had lesser confidence in these, so listing them separate from the above 'seem useful' ones
e9c5e7402dad6f4f04c2430db6f283512bcd4392 drm/amdgpu: add missing irq.h include => seems useful for building amdgpu on ARM
727f28b8ca24a581c7bd868326b8cea1058c720a tpm_tis: Use devm_free_irq not free_irq => simple fix
757647e10e55c01fb7a9c4356529442e316a7c72 hv_netvsc: use skb_get_hash() instead of a homegrown implementation => fixes performance regression
4aec6ec0da9c72c0fa1a5b0d1133707481347bb3 cxl: Increase timeout for detection of AFU mmio hang => helps detect mmio hang
725fc629ff2545b061407305ae51016c9f928fce kernek/fork.c: allocate idle task for a CPU always on its local node => performance improvement?
cd95aad5579371ac332507fc946008217fc37e6c Drivers: hv: vmbus: handle various crash scenarios => avoid kernel crashes
Hope these make some sense; with these, I've scanned through about 2300 patches till Jan'17 (Ubuntu-4.4.0-62.83 tag). I will continue to do this for the changes since then.
Best regards, Sumit
On Fri, Mar 17, 2017 at 6:59 PM, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
I've only started looking at OpenSUSE. There are a couple of interesting patches in the "stable" branch for 4.10, they all apply on top of v4.10.3:
24835e442f28 ("drm: reference count event->completion") 8aac7f343697 ("fbcon: Fix vc attr at deinit") 31c05415f5b4 ("bonding: use ETH_MAX_MTU as max mtu") 620711944459 ("crypto: algif_hash - avoid zero-sized array") d8e9b2b9097c ("drm/i915/gvt: Fix superfluous newline in GVT_DISPLAY_READY env var") dfcb9f4f99f1 ("sctp: deny peeloff operation on asocs with threads sleeping on it")
For 4.4, there are two interesting branches, SLE12-SP3 with 9354 patches and openSUSE-42.2 with 5502 patches. I'm sure I can find a lot in there, but have to come up with a little more automation as the approach of manually checking each patch as I did for the 4.10 branch probably won't scale here.
Arnd
On Fri, Mar 17, 2017 at 6:59 PM, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
I've only started looking at OpenSUSE. There are a couple of interesting patches in the "stable" branch for 4.10, they all apply on top of v4.10.3:
24835e442f28 ("drm: reference count event->completion") 8aac7f343697 ("fbcon: Fix vc attr at deinit") 31c05415f5b4 ("bonding: use ETH_MAX_MTU as max mtu") 620711944459 ("crypto: algif_hash - avoid zero-sized array") d8e9b2b9097c ("drm/i915/gvt: Fix superfluous newline in GVT_DISPLAY_READY env var") dfcb9f4f99f1 ("sctp: deny peeloff operation on asocs with threads sleeping on it")
For 4.4, there are two interesting branches, SLE12-SP3 with 9354 patches and openSUSE-42.2 with 5502 patches. I'm sure I can find a lot in there, but have to come up with a little more automation as the approach of manually checking each patch as I did for the 4.10 branch probably won't scale here.
Arnd
On Fri, Mar 17, 2017 at 10:52 PM, Arnd Bergmann arnd@linaro.org wrote:
On Fri, Mar 17, 2017 at 6:59 PM, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
I've only started looking at OpenSUSE. There are a couple of interesting patches in the "stable" branch for 4.10, they all apply on top of v4.10.3:
I've walked through the list again for v4.4 and v4.9:
24835e442f28 ("drm: reference count event->completion")
Applies cleanly to v4.9 but not v4.4, looks correct
8aac7f343697 ("fbcon: Fix vc attr at deinit")
Applies cleanly to v4.9+, looks correct
Rebasing to v4.4 requires a one-line fixup
31c05415f5b4 ("bonding: use ETH_MAX_MTU as max mtu")
Applies to v4.9 but is only needed on v4.10
620711944459 ("crypto: algif_hash - avoid zero-sized array")
Applies to v4.4+, obviously correct
d8e9b2b9097c ("drm/i915/gvt: Fix superfluous newline in GVT_DISPLAY_READY env var")
Applies only to v4.10
dfcb9f4f99f1 ("sctp: deny peeloff operation on asocs with threads sleeping on it")
Applies to v4.4+, but only makes sense on v4.10+
For 4.4, there are two interesting branches, SLE12-SP3 with 9354 patches and openSUSE-42.2 with 5502 patches. I'm sure I can find a lot in there, but have to come up with a little more automation as the approach of manually checking each patch as I did for the 4.10 branch probably won't scale here.
Still need to work my way through these
Arnd
On Wed, Mar 22, 2017 at 05:24:09PM +0100, Arnd Bergmann wrote:
On Fri, Mar 17, 2017 at 10:52 PM, Arnd Bergmann arnd@linaro.org wrote:
On Fri, Mar 17, 2017 at 6:59 PM, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
I've only started looking at OpenSUSE. There are a couple of interesting patches in the "stable" branch for 4.10, they all apply on top of v4.10.3:
I've walked through the list again for v4.4 and v4.9:
24835e442f28 ("drm: reference count event->completion")
Applies cleanly to v4.9 but not v4.4, looks correct
8aac7f343697 ("fbcon: Fix vc attr at deinit")
Applies cleanly to v4.9+, looks correct
Rebasing to v4.4 requires a one-line fixup
31c05415f5b4 ("bonding: use ETH_MAX_MTU as max mtu")
Applies to v4.9 but is only needed on v4.10
620711944459 ("crypto: algif_hash - avoid zero-sized array")
Applies to v4.4+, obviously correct
d8e9b2b9097c ("drm/i915/gvt: Fix superfluous newline in GVT_DISPLAY_READY env var")
Applies only to v4.10
dfcb9f4f99f1 ("sctp: deny peeloff operation on asocs with threads sleeping on it")
Applies to v4.4+, but only makes sense on v4.10+
Oops, missed these for this latest round, my fault. Will queue them up next week, thanks.
greg k-h
On Wed, Mar 22, 2017 at 05:24:09PM +0100, Arnd Bergmann wrote:
On Fri, Mar 17, 2017 at 10:52 PM, Arnd Bergmann arnd@linaro.org wrote:
On Fri, Mar 17, 2017 at 6:59 PM, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
I've only started looking at OpenSUSE. There are a couple of interesting patches in the "stable" branch for 4.10, they all apply on top of v4.10.3:
I've walked through the list again for v4.4 and v4.9:
24835e442f28 ("drm: reference count event->completion")
Applies cleanly to v4.9 but not v4.4, looks correct
8aac7f343697 ("fbcon: Fix vc attr at deinit")
Applies cleanly to v4.9+, looks correct
Rebasing to v4.4 requires a one-line fixup
31c05415f5b4 ("bonding: use ETH_MAX_MTU as max mtu")
Applies to v4.9 but is only needed on v4.10
620711944459 ("crypto: algif_hash - avoid zero-sized array")
Applies to v4.4+, obviously correct
d8e9b2b9097c ("drm/i915/gvt: Fix superfluous newline in GVT_DISPLAY_READY env var")
Applies only to v4.10
dfcb9f4f99f1 ("sctp: deny peeloff operation on asocs with threads sleeping on it")
Applies to v4.4+, but only makes sense on v4.10+
Thanks for these, all now applied.
greg k-h
On Fri, Mar 17, 2017 at 11:29:44PM +0530, Sumit Semwal wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
cc: stable ab2a4bf83902c170d29ba130a8abb5f9d90559e1 USB: don't free bandwidth_mutex too early
Doesn't apply.
c538b9436751a0be2e1246b48353bc23156bdbcc pinctrl: intel: Only restore pins that are used by the driver
I didn't apply this for some reason, are you sure it builds?
For things with a stable tag on it already, there has to be some reason I didn't apply it, either it needs to be fixed up or something else. If fixed up, can you provide the fixed patch?
I'll look at the rest of these on my next flight in a few hours...
thanks,
greg k-h
On Sat, Mar 18, 2017 at 10:55:08PM +0800, Greg KH wrote:
On Fri, Mar 17, 2017 at 11:29:44PM +0530, Sumit Semwal wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
cc: stable ab2a4bf83902c170d29ba130a8abb5f9d90559e1 USB: don't free bandwidth_mutex too early
Doesn't apply.
c538b9436751a0be2e1246b48353bc23156bdbcc pinctrl: intel: Only restore pins that are used by the driver
I didn't apply this for some reason, are you sure it builds?
Yup, it breaks the build as-is :(
Hi Greg,
On 18-Mar-2017 20:25, "Greg KH" gregkh@google.com wrote:
On Sat, Mar 18, 2017 at 10:55:08PM +0800, Greg KH wrote:
On Fri, Mar 17, 2017 at 11:29:44PM +0530, Sumit Semwal wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
cc: stable ab2a4bf83902c170d29ba130a8abb5f9d90559e1 USB: don't free bandwidth_mutex too early
Doesn't apply.
I'll try to resolve the issue and send.
c538b9436751a0be2e1246b48353bc23156bdbcc pinctrl: intel: Only restore pins that are used by the driver
I didn't apply this for some reason, are you sure it builds?
Yup, it breaks the build as-is :(
:( Sorry, my bad. For the list I sent you, haven't tried to apply and/or build them. I'll try to apply all these patches onto your tree, and sort them accordingly.
Hi Greg,
On 18 March 2017 at 20:30, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
On 18-Mar-2017 20:25, "Greg KH" gregkh@google.com wrote:
On Sat, Mar 18, 2017 at 10:55:08PM +0800, Greg KH wrote:
On Fri, Mar 17, 2017 at 11:29:44PM +0530, Sumit Semwal wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
cc: stable ab2a4bf83902c170d29ba130a8abb5f9d90559e1 USB: don't free bandwidth_mutex too early
Doesn't apply.
I'll try to resolve the issue and send.
This required another patch [1], also cc:stable which also didn't apply, so I resolved the minor conflict and then this applies. Am attaching the conflict-resolved patch here, if you'd like to take that. With these 2, the code builds fine, but I've got no way to test it.
c538b9436751a0be2e1246b48353bc23156bdbcc pinctrl: intel: Only restore pins that are used by the driver
I didn't apply this for some reason, are you sure it builds?
Yup, it breaks the build as-is :(
Yes :( - and the dependent commit isn't a 'clean' commit - [2]: added gpiochip_line_is_irq as part of changes to sunxi pinctrl. Best to drop this, perhaps.
:( Sorry, my bad. For the list I sent you, haven't tried to apply and/or build them. I'll try to apply all these patches onto your tree, and sort them accordingly.
Will spend sometime on Monday to validate the other patches that I've listed.
Best, Sumit
[1]: upstream commit feb26ac31a2a5cb88d86680d9a94916a6343e9e6: usb: core: hub: hub_port_init lock controller instead of bus [2]: upstream commit 6cee3821e4e4bd6e6cdf0870b6c72d455460bd39: gpio/pinctrl: sunxi: stop poking around in private vars
On Sat, Mar 18, 2017 at 10:30:17PM +0530, Sumit Semwal wrote:
Hi Greg,
On 18 March 2017 at 20:30, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
On 18-Mar-2017 20:25, "Greg KH" gregkh@google.com wrote:
On Sat, Mar 18, 2017 at 10:55:08PM +0800, Greg KH wrote:
On Fri, Mar 17, 2017 at 11:29:44PM +0530, Sumit Semwal wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
cc: stable ab2a4bf83902c170d29ba130a8abb5f9d90559e1 USB: don't free bandwidth_mutex too early
Doesn't apply.
I'll try to resolve the issue and send.
This required another patch [1], also cc:stable which also didn't apply, so I resolved the minor conflict and then this applies. Am attaching the conflict-resolved patch here, if you'd like to take that. With these 2, the code builds fine, but I've got no way to test it.
c538b9436751a0be2e1246b48353bc23156bdbcc pinctrl: intel: Only restore pins that are used by the driver
I didn't apply this for some reason, are you sure it builds?
Yup, it breaks the build as-is :(
Yes :( - and the dependent commit isn't a 'clean' commit - [2]: added gpiochip_line_is_irq as part of changes to sunxi pinctrl. Best to drop this, perhaps.
:( Sorry, my bad. For the list I sent you, haven't tried to apply and/or build them. I'll try to apply all these patches onto your tree, and sort them accordingly.
Will spend sometime on Monday to validate the other patches that I've listed.
I've applied these two USB patches now, thanks.
greg k-h
Hi Greg,
As a follow-up to the last email, I had an action to validate that the suggested patches apply and build atleast; here's the status on that:
On 17 March 2017 at 23:29, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
<snip>
Seem useful; similar patches already in stable 3a020a723c65eb8ffa7c237faca26521a024e582 crypto: ghash-clmulni - Fix load failure
Applies, and builds on top of 4.4.55. (x86)
1a07834024dfca5c4bed5de8f8714306e0a11836 crypto: cryptd - Assign statesize properly
Applies, and builds on top of 4.4.55. (x86/arm64)
ddef482420b1ba8ec45e6123a7e8d3f67b21e5e3 crypto: mcryptd - Fix load failure
Applies, and builds on top of 4.4.55. (x86/arm64)
Seem useful: 83430833b4d4a9c9b23964babbeb1f36450f8136 cxlflash: Increase cmd_per_lun for better throughput
Applies; I couldn't build as it is for PPC.
e34fbbac669de0b7fb7803929d0477f35f6e2833 ACPI / video: skip evaluating _DOD when it does not exist
Applies, and builds on top of 4.4.55. (x86/arm64)
bcb48cca23ec9852739e4a464307fa29515bbe48 pinctrl: cherryview: Do not mask all interrupts in probe
Applies, and builds on top of 4.4.55. (x86)
77c0c9735bc0ba5898e637a3a20d6bcb50e3f67d Drivers: hv: balloon: don't crash when memory is added in non-sorted order
Applies, and builds on top of 4.4.55. (x86)
a9f61ca793becabdefab03b77568d6c6f8c1bc79 Drivers: hv: avoid vfree() on crash
Applies, and builds on top of 4.4.55. (x86)
707e59ba494372a90d245f18b0c78982caa88e48 xen/qspinlock: Don't kick CPU if IRQ is not initialized
Applies, and builds on top of 4.4.55. (x86)
708e75a3ee750dce1072134e630d66c4e6eaf63c KVM: PPC: Book3S PR: Fix illegal opcode emulation
Applies; I couldn't build as it is for PPC.
dba599091c191d209b1499511a524ad9657c0e5a s390/pci: fix use after free in dma_init
Applies; I couldn't build as it is for s390.
A few more 'maybe's' that aren't marked cc: stable, but seemed useful; I had lesser confidence in these, so listing them separate from the above 'seem useful' ones
e9c5e7402dad6f4f04c2430db6f283512bcd4392 drm/amdgpu: add missing irq.h include => seems useful for building amdgpu on ARM
Applies, and builds on top of 4.4.55. (x86/arm64)
727f28b8ca24a581c7bd868326b8cea1058c720a tpm_tis: Use devm_free_irq not free_irq => simple fix
Applies, and builds on top of 4.4.55. (x86)
757647e10e55c01fb7a9c4356529442e316a7c72 hv_netvsc: use skb_get_hash() instead of a homegrown implementation => fixes performance regression
Applies, and builds on top of 4.4.55. (x86) - but you might want to drop it, since drivers/net/* stable patches have a separate path.
4aec6ec0da9c72c0fa1a5b0d1133707481347bb3 cxl: Increase timeout for detection of AFU mmio hang => helps detect mmio hang
Doesn't apply; will investigate if it's dependent on other patches.
725fc629ff2545b061407305ae51016c9f928fce kernek/fork.c: allocate idle task for a CPU always on its local node => performance improvement?
Applies, and builds on top of 4.4.55. (x86/arm64)
cd95aad5579371ac332507fc946008217fc37e6c Drivers: hv: vmbus: handle various crash scenarios => avoid kernel crashes
Doesn't apply; will investigate if it's dependent on other patches.
Hope these make some sense; with these, I've scanned through about 2300 patches till Jan'17 (Ubuntu-4.4.0-62.83 tag). I will continue to do this for the changes since then.
Best regards, Sumit
Best, Sumit.
On Wed, Mar 22, 2017 at 12:01:31AM +0530, Sumit Semwal wrote:
Hi Greg,
As a follow-up to the last email, I had an action to validate that the suggested patches apply and build atleast; here's the status on that:
On 17 March 2017 at 23:29, Sumit Semwal sumit.semwal@linaro.org wrote:
Hi Greg,
Apologies on the delay since last email; here are some patches from ubuntu xenial 4.4 for your consideration towards stable-4.4.y.
<snip> > > Seem useful; similar patches already in stable > 3a020a723c65eb8ffa7c237faca26521a024e582 crypto: ghash-clmulni - Fix > load failure Applies, and builds on top of 4.4.55. (x86)
1a07834024dfca5c4bed5de8f8714306e0a11836 crypto: cryptd - Assign statesize properly
Applies, and builds on top of 4.4.55. (x86/arm64)
ddef482420b1ba8ec45e6123a7e8d3f67b21e5e3 crypto: mcryptd - Fix load failure
Applies, and builds on top of 4.4.55. (x86/arm64)
Seem useful: 83430833b4d4a9c9b23964babbeb1f36450f8136 cxlflash: Increase cmd_per_lun for better throughput
Applies; I couldn't build as it is for PPC.
e34fbbac669de0b7fb7803929d0477f35f6e2833 ACPI / video: skip evaluating _DOD when it does not exist
Applies, and builds on top of 4.4.55. (x86/arm64)
bcb48cca23ec9852739e4a464307fa29515bbe48 pinctrl: cherryview: Do not mask all interrupts in probe
Applies, and builds on top of 4.4.55. (x86)
77c0c9735bc0ba5898e637a3a20d6bcb50e3f67d Drivers: hv: balloon: don't crash when memory is added in non-sorted order
Applies, and builds on top of 4.4.55. (x86)
a9f61ca793becabdefab03b77568d6c6f8c1bc79 Drivers: hv: avoid vfree() on crash
Applies, and builds on top of 4.4.55. (x86)
707e59ba494372a90d245f18b0c78982caa88e48 xen/qspinlock: Don't kick CPU if IRQ is not initialized
Applies, and builds on top of 4.4.55. (x86)
708e75a3ee750dce1072134e630d66c4e6eaf63c KVM: PPC: Book3S PR: Fix illegal opcode emulation
Applies; I couldn't build as it is for PPC.
dba599091c191d209b1499511a524ad9657c0e5a s390/pci: fix use after free in dma_init
Applies; I couldn't build as it is for s390.
A few more 'maybe's' that aren't marked cc: stable, but seemed useful; I had lesser confidence in these, so listing them separate from the above 'seem useful' ones
e9c5e7402dad6f4f04c2430db6f283512bcd4392 drm/amdgpu: add missing irq.h include => seems useful for building amdgpu on ARM
Applies, and builds on top of 4.4.55. (x86/arm64)
727f28b8ca24a581c7bd868326b8cea1058c720a tpm_tis: Use devm_free_irq not free_irq => simple fix
Applies, and builds on top of 4.4.55. (x86)
757647e10e55c01fb7a9c4356529442e316a7c72 hv_netvsc: use skb_get_hash() instead of a homegrown implementation => fixes performance regression
Applies, and builds on top of 4.4.55. (x86) - but you might want to drop it, since drivers/net/* stable patches have a separate path.
No, for stable patches, David gives me mboxes for the latest 2 kernels I maintain, for 4.4, I'm on my own here. So I've queued this up now.
Also I've applied all the others you mention here, and added you to the cc: for them.
thanks,
greg k-h
-
Arnd Bergmann -
Arnd Bergmann -
Greg KH -
Sumit Semwal