brcmfmac: keep wiphy during PCIe driver lifetime
cover brcmfmac: keep wiphy during PCIe driver lifetime
From: Rafał Miłecki <[email protected]>
Driver's main init/attach function brcmf_attach() was handling both:
wiphy allocation and driver initialization. It meant getting a new wiphy on every init (initial, resume, error recovery).
For supplicants/authenticators and Linux users it's move convenient to have the same wiphy over driver's lifetime. It allows e.g. automatic recovery after a firmware crash.
This patchset was tested on BCM4366 (PCIe) and BCM43430 (SDIO).
Right now only PCIe makes use of keeping the same wiphy. I got RPi Zero W so I'm planning to add firmware crash recovery & keep wiphy for SDIO in the future.
Rafał Miłecki (3):
brcmfmac: move "cfg80211_ops" pointer to another struct
brcmfmac: split brcmf_attach() and brcmf_detach() functions
brcmfmac: don't realloc wiphy during PCIe reset
.../broadcom/brcm80211/brcmfmac/bus.h | 4 +- .../broadcom/brcm80211/brcmfmac/cfg80211.c | 1 - .../broadcom/brcm80211/brcmfmac/cfg80211.h | 1 - .../broadcom/brcm80211/brcmfmac/core.c | 42 ++++++++++++++----- .../broadcom/brcm80211/brcmfmac/core.h | 1 + .../broadcom/brcm80211/brcmfmac/pcie.c | 13 +++++- .../broadcom/brcm80211/brcmfmac/sdio.c | 15 +++++-- .../broadcom/brcm80211/brcmfmac/usb.c | 34 ++++++++++++--- 8 files changed, 87 insertions(+), 24 deletions(-)
Signature · 1 line
2.21.0
PATCH 1/3 brcmfmac: move "cfg80211_ops" pointer to another struct
From: Rafał Miłecki <[email protected]>
This moves "ops" pointer from "struct brcmf_cfg80211_info" to the "struct brcmf_pub". This movement makes it possible to allocate wiphy without attaching cfg80211 (brcmf_cfg80211_attach()). It's required for later separation of wiphy allocation and driver initialization.
While at it fix also an unlikely memory leak in the brcmf_attach().
--- .../net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c | 1 - .../net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.h | 1 - drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c | 9 ++++++--- drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.h | 1 + 4 files changed, 7 insertions(+), 5 deletions(-)
drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c +0 -1
@@ -7210,7 +7210,6 @@ void brcmf_cfg80211_detach(struct brcmf_cfg80211_info *cfg)brcmf_pno_detach (cfg);brcmf_btcoex_detach (cfg);wiphy_unregister (cfg->wiphy); -kfree (cfg->ops);wl_deinit_priv (cfg);brcmf_free_wiphy (cfg->wiphy);kfree (cfg);
drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.h +0 -1
@@ -292,7 +292,6 @@ struct brcmf_cfg80211_wowl {*/ struct brcmf_cfg80211_info {struct wiphy *wiphy; -struct cfg80211_ops *ops;struct brcmf_cfg80211_conf *conf;struct brcmf_p2p_info p2p;struct brcmf_btcoex_info *btcoex;
drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c +6 -3
@@ -1224,12 +1224,15 @@ int brcmf_attach(struct device *dev, struct brcmf_mp_device *settings)return -ENOMEM ; wiphy =wiphy_new (ops,sizeof (*drvr)); -if (!wiphy) +if (!wiphy) { +kfree (ops);return -ENOMEM ; + }set_wiphy_dev (wiphy, dev); drvr =wiphy_priv (wiphy); drvr->wiphy = wiphy; + drvr->ops = ops;for (i =0 ; i <ARRAY_SIZE (drvr->if2bss); i++) drvr->if2bss[i] =BRCMF_BSSIDX_INVALID ;
@@ -1262,12 +1265,10 @@ int brcmf_attach(struct device *dev, struct brcmf_mp_device *settings) goto fail; } - drvr->config->ops = ops;return 0 ; fail:brcmf_detach (dev); -kfree (ops);return ret; }
@@ -1353,6 +1354,8 @@ void brcmf_detach(struct device *dev) bus_if->drvr =NULL ; +kfree (drvr->ops); +wiphy_free (drvr->wiphy); }
drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.h +1 -0
@@ -97,6 +97,7 @@ struct brcmf_pub {struct brcmf_bus *bus_if;struct brcmf_proto *proto;struct wiphy *wiphy; +struct cfg80211_ops *ops;struct brcmf_cfg80211_info *config;/* Internal brcmf items */
Signature · 1 line
2.21.0
Arend Van Spriel
maintainerRe: [PATCH 1/3] brcmfmac: move "cfg80211_ops" pointer to another struct
Quoted 8 lines · Rafał Miłecki
From: Rafał Miłecki <[email protected]>
This moves "ops" pointer from "struct
brcmf_cfg80211_info" to the "structbrcmf_pub". This movement makes it possible to allocate wiphy without attaching cfg80211 (brcmf_cfg80211_attach()). It's required for later separation of wiphy allocation and driver initialization.While at it fix also an unlikely memory leak in the
brcmf_attach().
Always good ;-)
I recall there is some fiddling with the callback ops in cfg80211.c. Is that broken by this reorg. Need to look into that.
Regards,
Arend
Re: [PATCH 1/3] brcmfmac: move "cfg80211_ops" pointer to another struct
Quoted 8 lines · Arend Van Spriel
Quoted 7 lines · Rafał Miłecki
From: Rafał Miłecki <[email protected]>
This moves "ops" pointer from "struct
brcmf_cfg80211_info" to the "structbrcmf_pub". This movement makes it possible to allocate wiphy without attaching cfg80211 (brcmf_cfg80211_attach()). It's required for later separation of wiphy allocation and driver initialization.
Quoted 6 lines
Quoted 1 line
While at it fix also an unlikely memory leak in the
brcmf_attach().Always good ;-)
I recall there is some fiddling with the callback ops in
cfg80211.c. Is that broken by this reorg. Need to look into that.
I don't see how this patch could break that. It still calls brcmf_cfg80211_get_ops() and passes settings as an argument.
--
Rafał
Re: [PATCH 1/3] brcmfmac: move "cfg80211_ops" pointer to another struct
In reply to Rafał MiłeckiQuoted 10 lines · Rafał Miłecki
From: Rafał Miłecki <[email protected]>
This moves "ops" pointer from "struct
brcmf_cfg80211_info" to the "structbrcmf_pub". This movement makes it possible to allocate wiphy without attaching cfg80211 (brcmf_cfg80211_attach()). It's required for later separation of wiphy allocation and driver initialization.While at it fix also an unlikely memory leak in the
brcmf_attach().Signed-off-by Rafał Miłecki <[email protected]>
3 patches applied to wireless-drivers-next.git, thanks.
ba76ff25ee64 brcmfmac: move "cfg80211_ops" pointer to another struct 450914c39f88 brcmfmac: split brcmf_attach() and brcmf_detach() functions a1f5aac1765a brcmfmac: don't realloc wiphy during PCIe reset
Signature · 3 lines
https://patchwork.kernel.org/patch/11127163/ https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
PATCH 2/3 brcmfmac: split brcmf_attach() and brcmf_detach() functions
In reply to Rafał MiłeckiFrom: Rafał Miłecki <[email protected]>
Move code allocating/freeing wiphy out of above functions. This will allow reinitializing the driver (e.g. on some error) without allocating a new wiphy.
--- .../broadcom/brcm80211/brcmfmac/bus.h | 4 ++- .../broadcom/brcm80211/brcmfmac/core.c | 33 ++++++++++++++---- .../broadcom/brcm80211/brcmfmac/pcie.c | 13 +++++-- .../broadcom/brcm80211/brcmfmac/sdio.c | 15 ++++++-- .../broadcom/brcm80211/brcmfmac/usb.c | 34 +++++++++++++++---- 5 files changed, 80 insertions(+), 19 deletions(-)
drivers/net/wireless/broadcom/brcm80211/brcmfmac/bus.h +3 -1
@@ -253,10 +253,12 @@ void brcmf_rx_frame(struct device *dev, struct sk_buff *rxp, bool handle_event);/* Receive async event packet from firmware. Callee disposes of rxp. */ void brcmf_rx_event (struct device *dev,struct sk_buff *rxp); +int brcmf_alloc (struct device *dev,struct brcmf_mp_device *settings);/* Indication from bus module regarding presence/insertion of dongle. */ -int brcmf_attach (struct device *dev,struct brcmf_mp_device *settings); +int brcmf_attach (struct device *dev);/* Indication from bus module regarding removal/absence of dongle */ void brcmf_detach (struct device *dev); +void brcmf_free (struct device *dev);/* Indication from bus module that dongle should be reset */ void brcmf_dev_reset (struct device *dev);/* Request from bus module to initiate a coredump */
drivers/net/wireless/broadcom/brcm80211/brcmfmac/core.c +26 -7
@@ -1209,13 +1209,11 @@ static int brcmf_bus_started(struct brcmf_pub *drvr, struct cfg80211_ops *ops)return ret; } -int brcmf_attach (struct device *dev,struct brcmf_mp_device *settings) +int brcmf_alloc (struct device *dev,struct brcmf_mp_device *settings) {struct wiphy *wiphy;struct cfg80211_ops *ops;struct brcmf_pub *drvr =NULL ; -int ret =0 ; -int i;brcmf_dbg (TRACE ,"Enter\n" );
@@ -1233,6 +1231,21 @@ int brcmf_attach(struct device *dev, struct brcmf_mp_device *settings) drvr =wiphy_priv (wiphy); drvr->wiphy = wiphy; drvr->ops = ops; + drvr->bus_if =dev_get_drvdata (dev); + drvr->bus_if->drvr = drvr; + drvr->settings = settings; + +return 0 ; +} + +int brcmf_attach (struct device *dev) +{ +struct brcmf_bus *bus_if =dev_get_drvdata (dev); +struct brcmf_pub *drvr = bus_if->drvr; +int ret =0 ; +int i; + +brcmf_dbg (TRACE ,"Enter\n" );for (i =0 ; i <ARRAY_SIZE (drvr->if2bss); i++) drvr->if2bss[i] =BRCMF_BSSIDX_INVALID ;
@@ -1241,9 +1254,6 @@ int brcmf_attach(struct device *dev, struct brcmf_mp_device *settings)/* Link to bus module */ drvr->hdrlen =0 ; - drvr->bus_if =dev_get_drvdata (dev); - drvr->bus_if->drvr = drvr; - drvr->settings = settings;/* Attach and link in the protocol */ ret =brcmf_proto_attach (drvr);
@@ -1259,7 +1269,7 @@ int brcmf_attach(struct device *dev, struct brcmf_mp_device *settings)/* attach firmware event handler */ brcmf_fweh_attach (drvr); - ret =brcmf_bus_started (drvr, ops); + ret =brcmf_bus_started (drvr, drvr->ops);if (ret !=0 ) {bphy_err (drvr,"dongle is not responding: err=%d\n" , ret); goto fail;
@@ -1351,6 +1361,15 @@ void brcmf_detach(struct device *dev)brcmf_cfg80211_detach (drvr->config); drvr->config =NULL ; } +} + +void brcmf_free (struct device *dev) +{ +struct brcmf_bus *bus_if =dev_get_drvdata (dev); +struct brcmf_pub *drvr = bus_if->drvr; + +if (!drvr) +return ; bus_if->drvr =NULL ;
drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c +11 -2
@@ -1430,6 +1430,7 @@ static int brcmf_pcie_reset(struct device *dev)brcmf_pcie_bus_console_read (devinfo, true);brcmf_detach (dev); +brcmf_free (dev);brcmf_pcie_release_irq (devinfo);brcmf_pcie_release_scratchbuffers (devinfo);
@@ -1824,11 +1825,18 @@ static void brcmf_pcie_setup(struct device *dev, int ret,brcmf_pcie_intr_enable (devinfo);brcmf_pcie_hostready (devinfo); -if (brcmf_attach (&devinfo->pdev->dev, devinfo->settings) ==0 ) -return ; + + ret =brcmf_alloc (&devinfo->pdev->dev, devinfo->settings); +if (ret) + goto fail; + ret =brcmf_attach (&devinfo->pdev->dev); +if (ret) + goto fail;brcmf_pcie_bus_console_read (devinfo, false); +return ; + fail:device_release_driver (dev); }
@@ -1971,6 +1979,7 @@ brcmf_pcie_remove(struct pci_dev *pdev)brcmf_pcie_intr_disable (devinfo);brcmf_detach (&pdev->dev); +brcmf_free (&pdev->dev);kfree (bus->bus_priv.pcie);kfree (bus->msgbuf->flowrings);
drivers/net/wireless/broadcom/brcm80211/brcmfmac/sdio.c +12 -3
@@ -4247,17 +4247,26 @@ static void brcmf_sdio_firmware_callback(struct device *dev, int err, sdiod->bus_if->chip = bus->ci->chip; sdiod->bus_if->chiprev = bus->ci->chiprev; + err =brcmf_alloc (sdiod->dev, sdiod->settings); +if (err) { +brcmf_err ("brcmf_alloc failed\n" ); + goto claim; + } +/* Attach to the common layer, reserve hdr space */ - err =brcmf_attach (sdiod->dev, sdiod->settings); + err =brcmf_attach (sdiod->dev);if (err !=0 ) {brcmf_err ("brcmf_attach failed\n" ); -sdio_claim_host (sdiod->func1); - goto checkdied; + goto free; }/* ready */ return ; +free: +brcmf_free (sdiod->dev); +claim: +sdio_claim_host (sdiod->func1); checkdied:brcmf_sdio_checkdied (bus); release:
drivers/net/wireless/broadcom/brcm80211/brcmfmac/usb.c +28 -6
@@ -1178,8 +1178,12 @@ static void brcmf_usb_probe_phase2(struct device *dev, int ret,if (ret) goto error; + ret =brcmf_alloc (devinfo->dev, devinfo->settings); +if (ret) + goto error; +/* Attach to the common driver interface */ - ret =brcmf_attach (devinfo->dev, devinfo->settings); + ret =brcmf_attach (devinfo->dev);if (ret) goto error;
@@ -1251,7 +1255,10 @@ static int brcmf_usb_probe_cb(struct brcmf_usbdev_info *devinfo) }if (!brcmf_usb_dlneeded (devinfo)) { - ret =brcmf_attach (devinfo->dev, devinfo->settings); + ret =brcmf_alloc (devinfo->dev, devinfo->settings); +if (ret) + goto fail; + ret =brcmf_attach (devinfo->dev);if (ret) goto fail;/* we are done */
@@ -1279,6 +1286,7 @@ static int brcmf_usb_probe_cb(struct brcmf_usbdev_info *devinfo) fail:/* Release resources in reverse order */ +brcmf_free (devinfo->dev);kfree (bus);brcmf_usb_detach (devinfo);return ret;
@@ -1292,6 +1300,7 @@ brcmf_usb_disconnect_cb(struct brcmf_usbdev_info *devinfo)brcmf_dbg (USB ,"Enter, bus_pub %p\n" , devinfo);brcmf_detach (devinfo->dev); +brcmf_free (devinfo->dev);kfree (devinfo->bus_pub.bus);brcmf_usb_detach (devinfo); }
@@ -1435,10 +1444,12 @@ static int brcmf_usb_suspend(struct usb_interface *intf, pm_message_t state)brcmf_dbg (USB ,"Enter\n" ); devinfo->bus_pub.state =BRCMFMAC_USB_STATE_SLEEP ; -if (devinfo->wowl_enabled) +if (devinfo->wowl_enabled) {brcmf_cancel_all_urbs (devinfo); -else + }else {brcmf_detach (&usb->dev); +brcmf_free (&usb->dev); + }return 0 ; }
@@ -1451,8 +1462,19 @@ static int brcmf_usb_resume(struct usb_interface *intf)struct brcmf_usbdev_info *devinfo =brcmf_usb_get_businfo (&usb->dev);brcmf_dbg (USB ,"Enter\n" ); -if (!devinfo->wowl_enabled) -return brcmf_attach (devinfo->dev, devinfo->settings); +if (!devinfo->wowl_enabled) { +int err; + + err =brcmf_alloc (&usb->dev, devinfo->settings); +if (err) +return err; + + err =brcmf_attach (devinfo->dev); +if (err) { +brcmf_free (devinfo->dev); +return err; + } + } devinfo->bus_pub.state =BRCMFMAC_USB_STATE_UP ;brcmf_usb_rx_fill_all (devinfo);
Signature · 1 line
2.21.0
Re: [PATCH 2/3] brcmfmac: split brcmf_attach() and brcmf_detach() functions
Quoted 5 lines · Rafał Miłecki
From: Rafał Miłecki <[email protected]>
Move code allocating/freeing wiphy out of above functions. This will allow reinitializing the driver (e.g. on some error) without allocating a new wiphy.
Quoted 8 lines
Signed-off-by Rafał Miłecki <[email protected]>--- .../broadcom/brcm80211/brcmfmac/bus.h | 4 ++- .../broadcom/brcm80211/brcmfmac/core.c | 33 ++++++++++++++---- .../broadcom/brcm80211/brcmfmac/pcie.c | 13 +++++-- .../broadcom/brcm80211/brcmfmac/sdio.c | 15 ++++++-- .../broadcom/brcm80211/brcmfmac/usb.c | 34 +++++++++++++++---- 5 files changed, 80 insertions(+), 19 deletions(-)
PATCH 3/3 brcmfmac: don't realloc wiphy during PCIe reset
In reply to Rafał MiłeckiFrom: Rafał Miłecki <[email protected]>
Providing a new wiphy on every PCIe reset was confusing and was causing configuration problems for some users (supplicant and authenticators). Sticking to the existing wiphy should make error recovery much simpler and more reliable.
--- drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c +4 -4
@@ -1430,7 +1430,6 @@ static int brcmf_pcie_reset(struct device *dev)brcmf_pcie_bus_console_read (devinfo, true);brcmf_detach (dev); -brcmf_free (dev);brcmf_pcie_release_irq (devinfo);brcmf_pcie_release_scratchbuffers (devinfo);
@@ -1826,9 +1825,6 @@ static void brcmf_pcie_setup(struct device *dev, int ret,brcmf_pcie_intr_enable (devinfo);brcmf_pcie_hostready (devinfo); - ret =brcmf_alloc (&devinfo->pdev->dev, devinfo->settings); -if (ret) - goto fail; ret =brcmf_attach (&devinfo->pdev->dev);if (ret) goto fail;
@@ -1931,6 +1927,10 @@ brcmf_pcie_probe(struct pci_dev *pdev, const struct pci_device_id *id) bus->wowl_supported =pci_pme_capable (pdev, PCI_D3hot);dev_set_drvdata (&pdev->dev, bus); + ret =brcmf_alloc (&devinfo->pdev->dev, devinfo->settings); +if (ret) + goto fail_bus; + fwreq =brcmf_pcie_prepare_fw_request (devinfo);if (!fwreq) { ret = -ENOMEM ;
Signature · 1 line
2.21.0
Re: [PATCH 3/3] brcmfmac: don't realloc wiphy during PCIe reset
Quoted 6 lines · Rafał Miłecki
From: Rafał Miłecki <[email protected]>
Providing a new wiphy on every PCIe reset was confusing and was causing configuration problems for some users (supplicant and authenticators). Sticking to the existing wiphy should make error recovery much simpler and more reliable.
Quoted 4 lines
Signed-off-by Rafał Miłecki <[email protected]>--- drivers/net/wireless/broadcom/brcm80211/brcmfmac/pcie.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
Arend Van Spriel
maintainerQuoted 9 lines · Rafał Miłecki
From: Rafał Miłecki <[email protected]>
Driver's main init/attach function
brcmf_attach()was handling both:
wiphy allocation and driver initialization. It meant getting a new wiphy on every init (initial, resume, error recovery).For supplicants/authenticators and Linux users it's move convenient to have the same wiphy over driver's lifetime. It allows e.g. automatic recovery after a firmware crash.
Typo: 'move' should be 'more'.
Regards,
Arend
Quoted 7 lines · Arend Van Spriel
Quoted 6 lines · Rafał Miłecki
From: Rafał Miłecki <[email protected]>
Driver's main init/attach function
brcmf_attach()was handling both:
wiphy allocation and driver initialization. It meant getting a new wiphy on every init (initial, resume, error recovery).
Quoted 5 lines
Quoted 3 lines
For supplicants/authenticators and Linux users it's move convenient to have the same wiphy over driver's lifetime. It allows e.g. automatic recovery after a firmware crash.
Typo: 'move' should be 'more'.
Well, it's just a cover letter (0/3), that message won't go anywhere to the git repository.
Signature · 1 line
Rafał