| Message ID | 171693842964.1298616.14265420982007939967.stgit@dwillia2-xfh.jf.intel.com |
|---|---|
| State | New, archived |
| Headers | show |
| Series | PCI: Make PCI cfg_access_lock lockdep key a singleton | expand |
On 5/28/24 4:22 PM, Dan Williams wrote: > The new lockdep annotation for cfg_access_lock naively registered a new > key per device. This is overkill and leads to warnings on hash > collisions at dynamic registration time: > > WARNING: CPU: 0 PID: 1 at kernel/locking/lockdep.c:1226 lockdep_register_key+0xb0/0x240 > RIP: 0010:lockdep_register_key+0xb0/0x240 > [..] > Call Trace: > <TASK> > ? __warn+0x8c/0x190 > ? lockdep_register_key+0xb0/0x240 > ? report_bug+0x1f8/0x200 > ? handle_bug+0x3c/0x70 > ? exc_invalid_op+0x18/0x70 > ? asm_exc_invalid_op+0x1a/0x20 > ? lockdep_register_key+0xb0/0x240 > pci_device_add+0x14b/0x560 > ? pci_setup_device+0x42e/0x6a0 > pci_scan_single_device+0xa7/0xd0 > p2sb_scan_and_cache_devfn+0xc/0x90 > p2sb_fs_init+0x15f/0x170 > > Switch to a shared static key for all instances. > > Fixes: 7e89efc6e9e4 ("PCI: Lock upstream bridge for pci_reset_function()") > Reported-by: Jani Saarinen <jani.saarinen@intel.com> > Closes: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_14834/bat-apl-1/boot0.txt > Cc: Dave Jiang <dave.jiang@intel.com> > Cc: Bjorn Helgaas <bhelgaas@google.com> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> Reviewed-by: Dave Jiang <dave.jiang@intel.com> > --- > Hi Bjorn, > > One more fallout from the cfg_access_lock lockdep annotation. This one > still wants a Tested-by from Jani before merging, but wanted to make you > aware of it in case similar reports make their way to you in the > meantime. > > drivers/pci/probe.c | 7 ++++--- > include/linux/pci.h | 1 - > 2 files changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c > index 8e696e547565..15168881ec94 100644 > --- a/drivers/pci/probe.c > +++ b/drivers/pci/probe.c > @@ -2533,6 +2533,8 @@ static void pci_set_msi_domain(struct pci_dev *dev) > dev_set_msi_domain(&dev->dev, d); > } > > +static struct lock_class_key cfg_access_key; > + > void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) > { > int ret; > @@ -2546,9 +2548,8 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) > dev->dev.dma_mask = &dev->dma_mask; > dev->dev.dma_parms = &dev->dma_parms; > dev->dev.coherent_dma_mask = 0xffffffffull; > - lockdep_register_key(&dev->cfg_access_key); > - lockdep_init_map(&dev->cfg_access_lock, dev_name(&dev->dev), > - &dev->cfg_access_key, 0); > + lockdep_init_map(&dev->cfg_access_lock, "&dev->cfg_access_lock", > + &cfg_access_key, 0); > > dma_set_max_seg_size(&dev->dev, 65536); > dma_set_seg_boundary(&dev->dev, 0xffffffff); > diff --git a/include/linux/pci.h b/include/linux/pci.h > index fb004fd4e889..5bece7fd11f8 100644 > --- a/include/linux/pci.h > +++ b/include/linux/pci.h > @@ -413,7 +413,6 @@ struct pci_dev { > struct resource driver_exclusive_resource; /* driver exclusive resource ranges */ > > bool match_driver; /* Skip attaching driver */ > - struct lock_class_key cfg_access_key; > struct lockdep_map cfg_access_lock; > > unsigned int transparent:1; /* Subtractive decode bridge */ >
On Tue, May 28, 2024 at 04:22:59PM -0700, Dan Williams wrote: > The new lockdep annotation for cfg_access_lock naively registered a new > key per device. This is overkill and leads to warnings on hash > collisions at dynamic registration time: > > WARNING: CPU: 0 PID: 1 at kernel/locking/lockdep.c:1226 lockdep_register_key+0xb0/0x240 > RIP: 0010:lockdep_register_key+0xb0/0x240 > [..] > Call Trace: > <TASK> > ? __warn+0x8c/0x190 > ? lockdep_register_key+0xb0/0x240 > ? report_bug+0x1f8/0x200 > ? handle_bug+0x3c/0x70 > ? exc_invalid_op+0x18/0x70 > ? asm_exc_invalid_op+0x1a/0x20 > ? lockdep_register_key+0xb0/0x240 > pci_device_add+0x14b/0x560 > ? pci_setup_device+0x42e/0x6a0 > pci_scan_single_device+0xa7/0xd0 > p2sb_scan_and_cache_devfn+0xc/0x90 > p2sb_fs_init+0x15f/0x170 > > Switch to a shared static key for all instances. > > Fixes: 7e89efc6e9e4 ("PCI: Lock upstream bridge for pci_reset_function()") > Reported-by: Jani Saarinen <jani.saarinen@intel.com> > Closes: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_14834/bat-apl-1/boot0.txt > Cc: Dave Jiang <dave.jiang@intel.com> > Cc: Bjorn Helgaas <bhelgaas@google.com> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> > --- Reviewed-by: Alison Schofield <alison.schofield@intel.com> >
Hi, > -----Original Message----- > From: Williams, Dan J <dan.j.williams@intel.com> > Sent: Wednesday, 29 May 2024 2.23 > To: bhelgaas@google.com > Cc: Saarinen, Jani <jani.saarinen@intel.com>; Jiang, Dave > <dave.jiang@intel.com>; linux-cxl@vger.kernel.org; linux-pci@vger.kernel.org > Subject: [PATCH] PCI: Make PCI cfg_access_lock lockdep key a singleton > > The new lockdep annotation for cfg_access_lock naively registered a new key > per device. This is overkill and leads to warnings on hash collisions at > dynamic registration time: > > WARNING: CPU: 0 PID: 1 at kernel/locking/lockdep.c:1226 > lockdep_register_key+0xb0/0x240 > RIP: 0010:lockdep_register_key+0xb0/0x240 > [..] > Call Trace: > <TASK> > ? __warn+0x8c/0x190 > ? lockdep_register_key+0xb0/0x240 > ? report_bug+0x1f8/0x200 > ? handle_bug+0x3c/0x70 > ? exc_invalid_op+0x18/0x70 > ? asm_exc_invalid_op+0x1a/0x20 > ? lockdep_register_key+0xb0/0x240 > pci_device_add+0x14b/0x560 > ? pci_setup_device+0x42e/0x6a0 > pci_scan_single_device+0xa7/0xd0 > p2sb_scan_and_cache_devfn+0xc/0x90 > p2sb_fs_init+0x15f/0x170 > > Switch to a shared static key for all instances. > > Fixes: 7e89efc6e9e4 ("PCI: Lock upstream bridge for pci_reset_function()") > Reported-by: Jani Saarinen <jani.saarinen@intel.com> > Closes: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_14834/bat-apl- > 1/boot0.txt > Cc: Dave Jiang <dave.jiang@intel.com> > Cc: Bjorn Helgaas <bhelgaas@google.com> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> > --- > Hi Bjorn, > > One more fallout from the cfg_access_lock lockdep annotation. This one still > wants a Tested-by from Jani before merging, but wanted to make you aware of > it in case similar reports make their way to you in the meantime. See patch sent to intel-gfx https://patchwork.freedesktop.org/series/134186/ => https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_134186v1/index.html?hosts=apl => No taint in the boot and system does testing again: https://intel-gfx-ci.01.org/tree/drm-tip/Patchwork_134186v1/bat-all.html?hosts=apl With that: Tested-By: Jani Saarinen <jani.saarinen@intel.com> Br, Jani > > drivers/pci/probe.c | 7 ++++--- > include/linux/pci.h | 1 - > 2 files changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c index > 8e696e547565..15168881ec94 100644 > --- a/drivers/pci/probe.c > +++ b/drivers/pci/probe.c > @@ -2533,6 +2533,8 @@ static void pci_set_msi_domain(struct pci_dev > *dev) > dev_set_msi_domain(&dev->dev, d); > } > > +static struct lock_class_key cfg_access_key; > + > void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) { > int ret; > @@ -2546,9 +2548,8 @@ void pci_device_add(struct pci_dev *dev, struct > pci_bus *bus) > dev->dev.dma_mask = &dev->dma_mask; > dev->dev.dma_parms = &dev->dma_parms; > dev->dev.coherent_dma_mask = 0xffffffffull; > - lockdep_register_key(&dev->cfg_access_key); > - lockdep_init_map(&dev->cfg_access_lock, dev_name(&dev->dev), > - &dev->cfg_access_key, 0); > + lockdep_init_map(&dev->cfg_access_lock, "&dev->cfg_access_lock", > + &cfg_access_key, 0); > > dma_set_max_seg_size(&dev->dev, 65536); > dma_set_seg_boundary(&dev->dev, 0xffffffff); diff --git > a/include/linux/pci.h b/include/linux/pci.h index fb004fd4e889..5bece7fd11f8 > 100644 > --- a/include/linux/pci.h > +++ b/include/linux/pci.h > @@ -413,7 +413,6 @@ struct pci_dev { > struct resource driver_exclusive_resource; /* driver exclusive > resource ranges */ > > bool match_driver; /* Skip attaching driver */ > - struct lock_class_key cfg_access_key; > struct lockdep_map cfg_access_lock; > > unsigned int transparent:1; /* Subtractive decode bridge > */
[+cc Alison, Imre, et al. IIUC this patch didn't help the similar issue reported by Imre at https://lore.kernel.org/r/ZlXP5oTnSApiDbD1@ideak-desk.fi.intel.com, but just FYI] On Tue, May 28, 2024 at 04:22:59PM -0700, Dan Williams wrote: > The new lockdep annotation for cfg_access_lock naively registered a new > key per device. This is overkill and leads to warnings on hash > collisions at dynamic registration time: > > WARNING: CPU: 0 PID: 1 at kernel/locking/lockdep.c:1226 lockdep_register_key+0xb0/0x240 > RIP: 0010:lockdep_register_key+0xb0/0x240 > [..] > Call Trace: > <TASK> > ? __warn+0x8c/0x190 > ? lockdep_register_key+0xb0/0x240 > ? report_bug+0x1f8/0x200 > ? handle_bug+0x3c/0x70 > ? exc_invalid_op+0x18/0x70 > ? asm_exc_invalid_op+0x1a/0x20 > ? lockdep_register_key+0xb0/0x240 > pci_device_add+0x14b/0x560 > ? pci_setup_device+0x42e/0x6a0 > pci_scan_single_device+0xa7/0xd0 > p2sb_scan_and_cache_devfn+0xc/0x90 > p2sb_fs_init+0x15f/0x170 > > Switch to a shared static key for all instances. > > Fixes: 7e89efc6e9e4 ("PCI: Lock upstream bridge for pci_reset_function()") > Reported-by: Jani Saarinen <jani.saarinen@intel.com> > Closes: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_14834/bat-apl-1/boot0.txt > Cc: Dave Jiang <dave.jiang@intel.com> > Cc: Bjorn Helgaas <bhelgaas@google.com> > Signed-off-by: Dan Williams <dan.j.williams@intel.com> Applied with Alison's reviewed-by and Jani's tested-by to for-linus for v6.10, thanks! > --- > Hi Bjorn, > > One more fallout from the cfg_access_lock lockdep annotation. This one > still wants a Tested-by from Jani before merging, but wanted to make you > aware of it in case similar reports make their way to you in the > meantime. > > drivers/pci/probe.c | 7 ++++--- > include/linux/pci.h | 1 - > 2 files changed, 4 insertions(+), 4 deletions(-) > > diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c > index 8e696e547565..15168881ec94 100644 > --- a/drivers/pci/probe.c > +++ b/drivers/pci/probe.c > @@ -2533,6 +2533,8 @@ static void pci_set_msi_domain(struct pci_dev *dev) > dev_set_msi_domain(&dev->dev, d); > } > > +static struct lock_class_key cfg_access_key; > + > void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) > { > int ret; > @@ -2546,9 +2548,8 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) > dev->dev.dma_mask = &dev->dma_mask; > dev->dev.dma_parms = &dev->dma_parms; > dev->dev.coherent_dma_mask = 0xffffffffull; > - lockdep_register_key(&dev->cfg_access_key); > - lockdep_init_map(&dev->cfg_access_lock, dev_name(&dev->dev), > - &dev->cfg_access_key, 0); > + lockdep_init_map(&dev->cfg_access_lock, "&dev->cfg_access_lock", > + &cfg_access_key, 0); > > dma_set_max_seg_size(&dev->dev, 65536); > dma_set_seg_boundary(&dev->dev, 0xffffffff); > diff --git a/include/linux/pci.h b/include/linux/pci.h > index fb004fd4e889..5bece7fd11f8 100644 > --- a/include/linux/pci.h > +++ b/include/linux/pci.h > @@ -413,7 +413,6 @@ struct pci_dev { > struct resource driver_exclusive_resource; /* driver exclusive resource ranges */ > > bool match_driver; /* Skip attaching driver */ > - struct lock_class_key cfg_access_key; > struct lockdep_map cfg_access_lock; > > unsigned int transparent:1; /* Subtractive decode bridge */ >
diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c index 8e696e547565..15168881ec94 100644 --- a/drivers/pci/probe.c +++ b/drivers/pci/probe.c @@ -2533,6 +2533,8 @@ static void pci_set_msi_domain(struct pci_dev *dev) dev_set_msi_domain(&dev->dev, d); } +static struct lock_class_key cfg_access_key; + void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) { int ret; @@ -2546,9 +2548,8 @@ void pci_device_add(struct pci_dev *dev, struct pci_bus *bus) dev->dev.dma_mask = &dev->dma_mask; dev->dev.dma_parms = &dev->dma_parms; dev->dev.coherent_dma_mask = 0xffffffffull; - lockdep_register_key(&dev->cfg_access_key); - lockdep_init_map(&dev->cfg_access_lock, dev_name(&dev->dev), - &dev->cfg_access_key, 0); + lockdep_init_map(&dev->cfg_access_lock, "&dev->cfg_access_lock", + &cfg_access_key, 0); dma_set_max_seg_size(&dev->dev, 65536); dma_set_seg_boundary(&dev->dev, 0xffffffff); diff --git a/include/linux/pci.h b/include/linux/pci.h index fb004fd4e889..5bece7fd11f8 100644 --- a/include/linux/pci.h +++ b/include/linux/pci.h @@ -413,7 +413,6 @@ struct pci_dev { struct resource driver_exclusive_resource; /* driver exclusive resource ranges */ bool match_driver; /* Skip attaching driver */ - struct lock_class_key cfg_access_key; struct lockdep_map cfg_access_lock; unsigned int transparent:1; /* Subtractive decode bridge */
The new lockdep annotation for cfg_access_lock naively registered a new key per device. This is overkill and leads to warnings on hash collisions at dynamic registration time: WARNING: CPU: 0 PID: 1 at kernel/locking/lockdep.c:1226 lockdep_register_key+0xb0/0x240 RIP: 0010:lockdep_register_key+0xb0/0x240 [..] Call Trace: <TASK> ? __warn+0x8c/0x190 ? lockdep_register_key+0xb0/0x240 ? report_bug+0x1f8/0x200 ? handle_bug+0x3c/0x70 ? exc_invalid_op+0x18/0x70 ? asm_exc_invalid_op+0x1a/0x20 ? lockdep_register_key+0xb0/0x240 pci_device_add+0x14b/0x560 ? pci_setup_device+0x42e/0x6a0 pci_scan_single_device+0xa7/0xd0 p2sb_scan_and_cache_devfn+0xc/0x90 p2sb_fs_init+0x15f/0x170 Switch to a shared static key for all instances. Fixes: 7e89efc6e9e4 ("PCI: Lock upstream bridge for pci_reset_function()") Reported-by: Jani Saarinen <jani.saarinen@intel.com> Closes: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_14834/bat-apl-1/boot0.txt Cc: Dave Jiang <dave.jiang@intel.com> Cc: Bjorn Helgaas <bhelgaas@google.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> --- Hi Bjorn, One more fallout from the cfg_access_lock lockdep annotation. This one still wants a Tested-by from Jani before merging, but wanted to make you aware of it in case similar reports make their way to you in the meantime. drivers/pci/probe.c | 7 ++++--- include/linux/pci.h | 1 - 2 files changed, 4 insertions(+), 4 deletions(-)