From: Bob Ham Subject: [PATCH 2/2] hwrng: chaoskey - Fix URB warning due to timeout on Alea Date: Fri, 3 Jun 2016 12:13:08 +0100 Message-ID: <1464952388-2405-3-git-send-email-bob.ham@collabora.com> References: <1464952388-2405-1-git-send-email-bob.ham@collabora.com> Cc: Matt Mackall , Herbert Xu , Clemens Ladisch , Greg KH To: linux-crypto@vger.kernel.org, linux-usb@vger.kernel.org, Keith Packard Return-path: Received: from verain.settrans.net ([93.93.131.233]:56424 "EHLO verain.settrans.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932675AbcFCLNa (ORCPT ); Fri, 3 Jun 2016 07:13:30 -0400 In-Reply-To: <1464952388-2405-1-git-send-email-bob.ham@collabora.com> Sender: linux-crypto-owner@vger.kernel.org List-ID: The first read on an Alea takes about 1.8 seconds, more than the timeout value waiting for the read. As a consequence, later URB reuse causes the warning given below. To avoid this, we increase the wait time for the first read on the Alea. [ 78.293247] WARNING: CPU: 3 PID: 1892 at drivers/usb/core/urb.c:338 usb_submit_urb+0x2b4/0x580 [usbcore] [ 78.293250] URB ffff8802135be3c0 submitted while active [ 78.293252] Modules linked in: chaoskey(+) rng_core rfcomm binfmt_misc bnep cfg80211 nfsd auth_rpcgss oid_registry nfs_acl nfs lockd grace fscache sunrpc bridge stp llc tun snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_codec_generic iTCO_wdt iTCO_vendor_support nls_utf8 nls_cp437 vfat fat intel_rapl x86_pkg_temp_thermal intel_powerclamp coretemp kvm_intel efi_pstore kvm irqbypass pcspkr btusb btrtl btbcm btintel uvcvideo joydev bluetooth videobuf2_vmalloc videobuf2_memops efivars videobuf2_v4l2 serio_raw i2c_i801 videobuf2_core videodev cdc_mbim media lpc_ich shpchp mfd_core cdc_ncm usbnet mii cdc_wdm cdc_acm evdev snd_hda_intel snd_hda_codec snd_hwdep snd_hda_core i915 snd_pcm snd_timer i2c_algo_bit drm_kms_helper wmi thinkpad_acpi drm nvram mei_me mei snd soundcore rfkill ac batter y i2c_core [ 78.293335] video button tpm_tis tpm fuse parport_pc ppdev lp parport autofs4 ext4 crc16 jbd2 mbcache algif_skcipher af_alg hid_generic usbhid hid dm_crypt dm_mod sg sr_mod cdrom sd_mod crct10dif_pclmul crc32_pclmul crc32c_intel jitterentropy_rng sha256_generic hmac drbg aesni_intel xhci_pci aes_x86_64 ahci glue_helper xhci_hcd ehci_pci lrw libahci gf128mul ablk_helper cryptd libata sdhci_pci psmouse sdhci scsi_mod ehci_hcd mmc_core usbcore usb_common thermal [ 78.293402] CPU: 3 PID: 1892 Comm: hwrng Not tainted 4.7.0-rc1-linux-14+ #16 [ 78.293405] Hardware name: LENOVO 232577G/232577G, BIOS G2ET92WW (2.52 ) 02/22/2013 [ 78.293408] 0000000000000000 ffffffff812dfa0f ffff8801fa5b3d68 0000000000000000 [ 78.293413] ffffffff81072224 ffff8802135be3c0 ffff8801fa5b3db8 ffff880212e44210 [ 78.293418] 0000000000000040 ffff880209fb32c0 ffff880212e44200 ffffffff8107228f [ 78.293422] Call Trace: [ 78.293432] [] ? dump_stack+0x5c/0x7d [ 78.293437] [] ? __warn+0xc4/0xe0 [ 78.293441] [] ? warn_slowpath_fmt+0x4f/0x60 [ 78.293451] [] ? enqueue_task_fair+0xcd2/0x1260 [ 78.293463] [] ? usb_submit_urb+0x2b4/0x580 [usbcore] [ 78.293474] [] ? __pm_runtime_resume+0x55/0x70 [ 78.293484] [] ? _chaoskey_fill+0x132/0x250 [chaoskey] [ 78.293485] usbcore: registered new interface driver chaoskey [ 78.293493] [] ? wait_woken+0x90/0x90 [ 78.293500] [] ? devm_hwrng_register+0x80/0x80 [rng_core] [ 78.293505] [] ? chaoskey_rng_read+0x127/0x140 [chaoskey] [ 78.293511] [] ? devm_hwrng_register+0x80/0x80 [rng_core] [ 78.293515] [] ? hwrng_fillfn+0x6e/0x120 [rng_core] [ 78.293520] [] ? kthread+0xcf/0xf0 [ 78.293529] [] ? ret_from_fork+0x1f/0x40 [ 78.293535] [] ? kthread_park+0x50/0x50 Signed-off-by: Bob Ham --- drivers/usb/misc/chaoskey.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-) diff --git a/drivers/usb/misc/chaoskey.c b/drivers/usb/misc/chaoskey.c index 9aef46b..6ddd08a 100644 --- a/drivers/usb/misc/chaoskey.c +++ b/drivers/usb/misc/chaoskey.c @@ -60,7 +60,8 @@ MODULE_LICENSE("GPL"); #define CHAOSKEY_BUF_LEN 64 /* max size of USB full speed packet */ -#define NAK_TIMEOUT (HZ) /* stall/wait timeout for device */ +#define NAK_TIMEOUT (HZ) /* normal stall/wait timeout */ +#define ALEA_FIRST_TIMEOUT (HZ*3) /* first stall/wait timeout for Alea */ #ifdef CONFIG_USB_DYNAMIC_MINORS #define USB_CHAOSKEY_MINOR_BASE 0 @@ -88,6 +89,7 @@ struct chaoskey { int open; /* open count */ bool present; /* device not disconnected */ bool reading; /* ongoing IO */ + bool reads_started; /* track first read for Alea */ int size; /* size of buf */ int valid; /* bytes of buf read */ int used; /* bytes of buf consumed */ @@ -192,6 +194,9 @@ static int chaoskey_probe(struct usb_interface *interface, dev->in_ep = in_ep; + if (udev->descriptor.idVendor != ALEA_VENDOR_ID) + dev->reads_started = 1; + dev->size = size; dev->present = 1; @@ -361,6 +366,7 @@ static int _chaoskey_fill(struct chaoskey *dev) { DEFINE_WAIT(wait); int result; + bool started; usb_dbg(dev->interface, "fill"); @@ -393,10 +399,17 @@ static int _chaoskey_fill(struct chaoskey *dev) goto out; } + /* The first read on the Alea takes a little under 2 seconds. + * Reads after the first read take only a few microseconds + * though. Presumably the entropy-generating circuit needs + * time to ramp up. So, we wait longer on the first read. + */ + started = dev->reads_started; + dev->reads_started = true; result = wait_event_interruptible_timeout( dev->wait_q, !dev->reading, - NAK_TIMEOUT); + (started ? NAK_TIMEOUT : ALEA_FIRST_TIMEOUT) ); if (result < 0) goto out; -- 2.1.4