Return-path: Received: from mtaout02-winn.ispmail.ntl.com ([81.103.221.48]:40468 "EHLO mtaout02-winn.ispmail.ntl.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757785Ab0J2N5h (ORCPT ); Fri, 29 Oct 2010 09:57:37 -0400 From: Daniel Drake To: linville@tuxdriver.com Cc: linux-wireless@vger.kernel.org Cc: dcbw@redhat.com Cc: libertas-dev@lists.infradead.org Cc: pgf@laptop.org Subject: [PATCH v3] libertas: Fix sd8686 firmware reload Message-Id: <20101029135728.DF4C99D401B@zog.reactivated.net> Date: Fri, 29 Oct 2010 14:57:28 +0100 (BST) Sender: linux-wireless-owner@vger.kernel.org List-ID: From: Paul Fox For the SD8686, we cannot rely on the scratch register to read the firmware load status, because the same register is used for storing RX packet length. Broaden the check to account for this. The module can now be unloaded/reloaded successfully. Based on the implementation from libertas_tf. Signed-off-by: Daniel Drake Acked-by: Dan Williams Signed-off-by: Steve deRosier --- drivers/net/wireless/libertas/if_sdio.c | 32 ++++++++++++++++++++++++++++-- 1 files changed, 29 insertions(+), 3 deletions(-) v2: rebase on latest pull to overcome a trivial conflict generated by another recent change v3: this one actually compiles, sorry about the last one diff --git a/drivers/net/wireless/libertas/if_sdio.c b/drivers/net/wireless/libertas/if_sdio.c index 296fd00..e5685dc 100644 --- a/drivers/net/wireless/libertas/if_sdio.c +++ b/drivers/net/wireless/libertas/if_sdio.c @@ -684,18 +684,40 @@ static int if_sdio_prog_firmware(struct if_sdio_card *card) lbs_deb_enter(LBS_DEB_SDIO); + /* + * Disable interrupts + */ + sdio_claim_host(card->func); + sdio_writeb(card->func, 0x00, IF_SDIO_H_INT_MASK, &ret); + sdio_release_host(card->func); + sdio_claim_host(card->func); scratch = if_sdio_read_scratch(card, &ret); sdio_release_host(card->func); + lbs_deb_sdio("firmware status = %#x\n", scratch); + lbs_deb_sdio("scratch ret = %d\n", ret); + if (ret) goto out; - lbs_deb_sdio("firmware status = %#x\n", scratch); + /* + * The manual clearly describes that FEDC is the right code to use + * to detect firmware presence, but for SD8686 it is not that simple. + * Scratch is also used to store the RX packet length, so we lose + * the FEDC value early on. So we use a non-zero check in order + * to validate firmware presence. + * Additionally, the SD8686 in the Gumstix always has the high scratch + * bit set, even when the firmware is not loaded. So we have to + * exclude that from the test. + */ if (scratch == IF_SDIO_FIRMWARE_OK) { lbs_deb_sdio("firmware already loaded\n"); goto success; + } else if ((card->model == MODEL_8686) && (scratch & 0x7fff)) { + lbs_deb_sdio("firmware may be running\n"); + goto success; } ret = lbs_get_firmware(&card->func->dev, lbs_helper_name, lbs_fw_name, @@ -709,10 +731,14 @@ static int if_sdio_prog_firmware(struct if_sdio_card *card) if (ret) goto out; + lbs_deb_sdio("Helper firmware loaded\n"); + ret = if_sdio_prog_real(card, mainfw); if (ret) goto out; + lbs_deb_sdio("Firmware loaded\n"); + success: sdio_claim_host(card->func); sdio_set_block_size(card->func, IF_SDIO_BLOCK_SIZE); @@ -1042,8 +1068,6 @@ static int if_sdio_probe(struct sdio_func *func, priv->exit_deep_sleep = if_sdio_exit_deep_sleep; priv->reset_deep_sleep_wakeup = if_sdio_reset_deep_sleep_wakeup; - priv->fw_ready = 1; - sdio_claim_host(func); /* @@ -1064,6 +1088,8 @@ static int if_sdio_probe(struct sdio_func *func, if (ret) goto reclaim; + priv->fw_ready = 1; + /* * FUNC_INIT is required for SD8688 WLAN/BT multiple functions */ -- 1.7.2.3