RE: [PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case
> -Original Message- > From: Bjorn Helgaas [mailto:helg...@kernel.org] > Sent: Wednesday, November 16, 2016 2:49 PM > To: Dexuan Cui <de...@microsoft.com> > Cc: Bjorn Helgaas <bhelg...@google.com>; linux-...@vger.kernel.org; > de...@linuxdriverproject.org; gre...@linuxfoundation.org; KY Srinivasan > <k...@microsoft.com>; Haiyang Zhang <haiya...@microsoft.com>; Stephen > Hemminger <sthem...@microsoft.com>; Jake Oshins > <ja...@microsoft.com>; Hadden Hoppert <hadd...@microsoft.com>; > Vitaly Kuznetsov <vkuzn...@redhat.com>; jasow...@redhat.com; > a...@canonical.com; o...@aepfle.de; linux-kernel@vger.kernel.org > Subject: Re: [PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case > > On Thu, Nov 10, 2016 at 07:16:22AM +, Dexuan Cui wrote: > > PATCH 1 is just a clean-up. There should be no functional change. > > > > PATCH 2 and 3 are for device hot-remove case. > > Currently the driver will stop working or even cause panic, if we do > > hot add/remove quickly a few times. With the 2 patches, everything works > > reliably in my tests now. > > > > There can be still a potential issue with hot-remove when we unload > > the driver at the same time. That would require more work of proper > > synchronization among the 3 paths: the .probe/.remove, the channel > callback, > > and the offloaded hv_pci_devices_present()/hv_eject_device_work(). > > > > But for now, PATCH 2 and 3 do improve the situation a lot. > > > > Dexuan Cui (3): > > PCI: hv: use the correct buffer size in new_pcichild_device() > > PCI: hv: fix hv_pci_remove() for hot-remove > > PCI: hv: delete the device earlier from hbus->children for hot-remove > > > > drivers/pci/host/pci-hyperv.c | 67 ++--- > -- > > 1 file changed, 40 insertions(+), 27 deletions(-) > > I applied all three of these to pci/host-hv for v4.10, thanks! > > Jake, I converted your "looks good to me" to Reviewed-by tags. > > K. Y., I added your acks on 2 & 3. If you acked 1, I missed it. Thanks Bjorn; I thought I had acked all three, but that is ok. K. Y
RE: [PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case
> -Original Message- > From: Bjorn Helgaas [mailto:helg...@kernel.org] > Sent: Wednesday, November 16, 2016 2:49 PM > To: Dexuan Cui > Cc: Bjorn Helgaas ; linux-...@vger.kernel.org; > de...@linuxdriverproject.org; gre...@linuxfoundation.org; KY Srinivasan > ; Haiyang Zhang ; Stephen > Hemminger ; Jake Oshins > ; Hadden Hoppert ; > Vitaly Kuznetsov ; jasow...@redhat.com; > a...@canonical.com; o...@aepfle.de; linux-kernel@vger.kernel.org > Subject: Re: [PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case > > On Thu, Nov 10, 2016 at 07:16:22AM +, Dexuan Cui wrote: > > PATCH 1 is just a clean-up. There should be no functional change. > > > > PATCH 2 and 3 are for device hot-remove case. > > Currently the driver will stop working or even cause panic, if we do > > hot add/remove quickly a few times. With the 2 patches, everything works > > reliably in my tests now. > > > > There can be still a potential issue with hot-remove when we unload > > the driver at the same time. That would require more work of proper > > synchronization among the 3 paths: the .probe/.remove, the channel > callback, > > and the offloaded hv_pci_devices_present()/hv_eject_device_work(). > > > > But for now, PATCH 2 and 3 do improve the situation a lot. > > > > Dexuan Cui (3): > > PCI: hv: use the correct buffer size in new_pcichild_device() > > PCI: hv: fix hv_pci_remove() for hot-remove > > PCI: hv: delete the device earlier from hbus->children for hot-remove > > > > drivers/pci/host/pci-hyperv.c | 67 ++--- > -- > > 1 file changed, 40 insertions(+), 27 deletions(-) > > I applied all three of these to pci/host-hv for v4.10, thanks! > > Jake, I converted your "looks good to me" to Reviewed-by tags. > > K. Y., I added your acks on 2 & 3. If you acked 1, I missed it. Thanks Bjorn; I thought I had acked all three, but that is ok. K. Y
Re: [PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case
On Thu, Nov 10, 2016 at 07:16:22AM +, Dexuan Cui wrote: > PATCH 1 is just a clean-up. There should be no functional change. > > PATCH 2 and 3 are for device hot-remove case. > Currently the driver will stop working or even cause panic, if we do > hot add/remove quickly a few times. With the 2 patches, everything works > reliably in my tests now. > > There can be still a potential issue with hot-remove when we unload > the driver at the same time. That would require more work of proper > synchronization among the 3 paths: the .probe/.remove, the channel callback, > and the offloaded hv_pci_devices_present()/hv_eject_device_work(). > > But for now, PATCH 2 and 3 do improve the situation a lot. > > Dexuan Cui (3): > PCI: hv: use the correct buffer size in new_pcichild_device() > PCI: hv: fix hv_pci_remove() for hot-remove > PCI: hv: delete the device earlier from hbus->children for hot-remove > > drivers/pci/host/pci-hyperv.c | 67 > ++- > 1 file changed, 40 insertions(+), 27 deletions(-) I applied all three of these to pci/host-hv for v4.10, thanks! Jake, I converted your "looks good to me" to Reviewed-by tags. K. Y., I added your acks on 2 & 3. If you acked 1, I missed it.
Re: [PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case
On Thu, Nov 10, 2016 at 07:16:22AM +, Dexuan Cui wrote: > PATCH 1 is just a clean-up. There should be no functional change. > > PATCH 2 and 3 are for device hot-remove case. > Currently the driver will stop working or even cause panic, if we do > hot add/remove quickly a few times. With the 2 patches, everything works > reliably in my tests now. > > There can be still a potential issue with hot-remove when we unload > the driver at the same time. That would require more work of proper > synchronization among the 3 paths: the .probe/.remove, the channel callback, > and the offloaded hv_pci_devices_present()/hv_eject_device_work(). > > But for now, PATCH 2 and 3 do improve the situation a lot. > > Dexuan Cui (3): > PCI: hv: use the correct buffer size in new_pcichild_device() > PCI: hv: fix hv_pci_remove() for hot-remove > PCI: hv: delete the device earlier from hbus->children for hot-remove > > drivers/pci/host/pci-hyperv.c | 67 > ++- > 1 file changed, 40 insertions(+), 27 deletions(-) I applied all three of these to pci/host-hv for v4.10, thanks! Jake, I converted your "looks good to me" to Reviewed-by tags. K. Y., I added your acks on 2 & 3. If you acked 1, I missed it.
[PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case
PATCH 1 is just a clean-up. There should be no functional change. PATCH 2 and 3 are for device hot-remove case. Currently the driver will stop working or even cause panic, if we do hot add/remove quickly a few times. With the 2 patches, everything works reliably in my tests now. There can be still a potential issue with hot-remove when we unload the driver at the same time. That would require more work of proper synchronization among the 3 paths: the .probe/.remove, the channel callback, and the offloaded hv_pci_devices_present()/hv_eject_device_work(). But for now, PATCH 2 and 3 do improve the situation a lot. Dexuan Cui (3): PCI: hv: use the correct buffer size in new_pcichild_device() PCI: hv: fix hv_pci_remove() for hot-remove PCI: hv: delete the device earlier from hbus->children for hot-remove drivers/pci/host/pci-hyperv.c | 67 ++- 1 file changed, 40 insertions(+), 27 deletions(-) -- 2.7.4
[PATCH 0/3] PCI: hv: clean-up and 2 fixes to the hot-remove case
PATCH 1 is just a clean-up. There should be no functional change. PATCH 2 and 3 are for device hot-remove case. Currently the driver will stop working or even cause panic, if we do hot add/remove quickly a few times. With the 2 patches, everything works reliably in my tests now. There can be still a potential issue with hot-remove when we unload the driver at the same time. That would require more work of proper synchronization among the 3 paths: the .probe/.remove, the channel callback, and the offloaded hv_pci_devices_present()/hv_eject_device_work(). But for now, PATCH 2 and 3 do improve the situation a lot. Dexuan Cui (3): PCI: hv: use the correct buffer size in new_pcichild_device() PCI: hv: fix hv_pci_remove() for hot-remove PCI: hv: delete the device earlier from hbus->children for hot-remove drivers/pci/host/pci-hyperv.c | 67 ++- 1 file changed, 40 insertions(+), 27 deletions(-) -- 2.7.4
