Ricardo,
while looking into my performance issue I noticed that the PCI bridge in the 3380 reports a negotiated connection speed of 2.5GT/s (instead of the expected 5GT/s)?
The abbreviated lspci -vv output on one of my test machine(s):
01:00.0 PCI bridge: PLX Technology, Inc. Device 3380 (rev ab) (prog-if 00 [Normal decode])
LnkCap: Port #0, Speed 5GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <1us, L1 <2us
LnkSta: Speed 5GT/s, Width x1, TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
02:02.0 PCI bridge: PLX Technology, Inc. Device 3380 (rev ab) (prog-if 00 [Normal decode])
LnkCap: Port #2, Speed 5GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <2us, L1 <4us
LnkSta: Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
03:00.0 USB controller: PLX Technology, Inc. Device 3380 (rev ab) (prog-if fe [USB Device])
LnkCap: Port #3, Speed 5GT/s, Width x1, ASPM L0s L1, Exit Latency L0s <2us, L1 <4us
LnkSta: Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk- DLActive- BWMgmt- ABWMgmt-
The above appears to be the same on 3.17.0, 3.19.16 and 4.0.1 kernels.
If the above is accurate it would at least explain a factor 2 in performance.
I tried switching of PCI express power management by passing pcie_aspm=off to the kernel at startup, but that didn’t change anything.
Any ideas on how to force the internal connection to 5GT/s?
Regards,
Paul.
On 18 Jan 2015, at 16:49, Paul Jones <p.jones@xxxxxxxxxx> wrote:
> Ricardo,
>
> I think I figured out the problem: my 3380 was running in legacy adapter mode.
> I am now capable of connecting both g_mass_storage and g_ether without any kernel panics after ensuring the 3380 is in enhanced adapter mode.
>
> My only concern is the speed from my Mac to Linux:
> - g_ether: scp transfer 42Mb/s
> - g_mass_storage: dd bs=64k 116Mb/s (backed by SSD storage)
> For comparison, a direct gigabit connection allows between 90 and 95Mb/s using scp between the machines.
> Local writes on the linux side using dd is around 480Mb/s (on the same SSD storage).
>
> Any ideas on how to achieve a higher performance?
>
> Regards,
> Paul.
>
> On 25 Nov 2014, at 16:30, Paul Jones <p.jones@xxxxxxxxxx> wrote:
>
>> Ricardo,
>>
>> Unfortunately your latest change gives similar results, cycling errors in the log:
>> [ 201.287706] ------------[ cut here ]------------
>> [ 201.288328] WARNING: CPU: 3 PID: 1500 at drivers/usb/gadget/udc/net2280.c:816 start_dma+0x153/0x160 [net2280]()
>> [ 201.288991] Modules linked in: g_mass_storage usb_f_mass_storage libcomposite configfs net2280 udc_core snd_hda_codec_hdmi joydev hid_generic usbhid hid i915 intel_rapl x86_pkg_temp_thermal intel_powerclamp snd_hda_codec_realtek coretemp kvm_intel snd_hda_codec_generic kvm snd_hda_intel eeepc_wmi asus_wmi snd_hda_controller rfcomm snd_seq_midi bnep snd_seq_midi_event bluetooth sparse_keymap snd_hda_codec crct10dif_pclmul snd_rawmidi snd_hwdep crc32_pclmul snd_pcm ghash_clmulni_intel aesni_intel aes_x86_64 drm_kms_helper lrw gf128mul glue_helper ablk_helper cryptd snd_seq mei_me mei serio_raw drm lpc_ich video mac_hid tpm_infineon wmi snd_seq_device snd_timer snd parport_pc i2c_algo_bit soundcore ppdev lp parport e1000e psmouse r8169 ahci libahci mii ptp pps_core
>> [ 201.292574] CPU: 3 PID: 1500 Comm: file-storage Tainted: G W 3.17.0-rc5+ #2
>> [ 201.293277] Hardware name: ASUS All Series/Q87T, BIOS 0215 09/06/2013
>> [ 201.293976] 0000000000000009 ffff8803f1d8fd28 ffffffff81746707 0000000000000000
>> [ 201.294689] ffff8803f1d8fd60 ffffffff8106c93d ffff8803efc26478 ffff8800ce317100
>> [ 201.295398] ffffc90005b9c1a0 ffff8800ce317168 ffff8803efc26320 ffff8803f1d8fd70
>> [ 201.296109] Call Trace:
>> [ 201.296815] [<ffffffff81746707>] dump_stack+0x45/0x56
>> [ 201.297524] [<ffffffff8106c93d>] warn_slowpath_common+0x7d/0xa0
>> [ 201.298218] [<ffffffff8106ca1a>] warn_slowpath_null+0x1a/0x20
>> [ 201.298891] [<ffffffffa0428103>] start_dma+0x153/0x160 [net2280]
>> [ 201.299554] [<ffffffffa0429b1b>] net2280_queue+0x2db/0x480 [net2280]
>> [ 201.300209] [<ffffffffa04724f1>] start_transfer.isra.32+0x71/0xe0 [usb_f_mass_storage]
>> [ 201.300851] [<ffffffffa047259e>] start_out_transfer+0x3e/0x80 [usb_f_mass_storage]
>> [ 201.301474] [<ffffffffa0473967>] fsg_main_thread+0x207/0x17f0 [usb_f_mass_storage]
>> [ 201.302069] [<ffffffff81749eda>] ? __schedule+0x37a/0x830
>> [ 201.302607] [<ffffffffa0473760>] ? handle_exception+0x460/0x460 [usb_f_mass_storage]
>> [ 201.303200] [<ffffffff8108a3e2>] kthread+0xd2/0xf0
>> [ 201.303776] [<ffffffff8108a310>] ? kthread_create_on_node+0x180/0x180
>> [ 201.304328] [<ffffffff8174ecbc>] ret_from_fork+0x7c/0xb0
>> [ 201.304973] [<ffffffff8108a310>] ? kthread_create_on_node+0x180/0x180
>> [ 201.305486] ---[ end trace a7f3e86a1a37203b ]—
>> Followed by:
>> [ 263.311338] net2280 0000:01:00.0: The dmastat return = 5002!!
>> [ 263.409818] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
>>
>> as long as you have ideas, I’ll be more than happy to try them :)
>>
>> Paul.
>>
>> On 25 Nov 2014, at 15:59, Ricardo Ribalda Delgado <ricardo.ribalda@xxxxxxxxx> wrote:
>>
>>> One last try :)
>>>
>>> Instead of:
>>>
>>> if (likely(t & BIT(FIFO_EMPTY))) {
>>>
>>> have this:
>>>
>>> if ( t & BIT(NAK_OUT_PACKETS)){
>>> count = readl(&ep->dma->dmacount);
>>> count &= DMA_BYTE_COUNT_MASK;
>>> break;
>>> }
>>>
>>> if (likely(t & BIT(FIFO_EMPTY))) {
>>>
>>>
>>>
>>> On Tue, Nov 25, 2014 at 3:54 PM, Paul Jones <p.jones@xxxxxxxxxx> wrote:
>>>> Ricardo,
>>>>
>>>> it no longer locks up but if I try to write to the drive, I get cycles of:
>>>> [ 2334.127653] ------------[ cut here ]------------
>>>> [ 2334.128318] WARNING: CPU: 0 PID: 2140 at drivers/usb/gadget/udc/net2280.c:816 start_dma+0x153/0x160 [net2280]()
>>>> [ 2334.129105] Modules linked in: g_mass_storage usb_f_mass_storage libcomposite configfs net2280 udc_core snd_hda_codec_hdmi i915 rfcomm bnep bluetooth snd_hda_codec_realtek snd_hda_codec_generic intel_rapl x86_pkg_temp_thermal snd_hda_intel intel_powerclamp coretemp snd_hda_controller kvm_intel snd_hda_codec kvm drm_kms_helper snd_hwdep snd_pcm drm snd_seq_midi joydev snd_seq_midi_event snd_rawmidi hid_generic snd_seq eeepc_wmi asus_wmi sparse_keymap crct10dif_pclmul usbhid crc32_pclmul ghash_clmulni_intel aesni_intel snd_seq_device snd_timer aes_x86_64 hid lrw gf128mul glue_helper snd ablk_helper mei_me mei cryptd serio_raw lpc_ich i2c_algo_bit wmi soundcore video tpm_infineon mac_hid parport_pc ppdev lp parport e1000e psmouse r8169 ahci ptp libahci mii pps_core
>>>> [ 2334.132731] CPU: 0 PID: 2140 Comm: file-storage Tainted: G W 3.17.0-rc5+ #2
>>>> [ 2334.133438] Hardware name: ASUS All Series/Q87T, BIOS 0215 09/06/2013
>>>> [ 2334.134146] 0000000000000009 ffff8804087a3d28 ffffffff81746707 0000000000000000
>>>> [ 2334.134863] ffff8804087a3d60 ffffffff8106c93d ffff88002e693478 ffff8803efd14480
>>>> [ 2334.135583] ffffc900018fc1a0 ffff8803efd144e8 ffff88002e693320 ffff8804087a3d70
>>>> [ 2334.136299] Call Trace:
>>>> [ 2334.137023] [<ffffffff81746707>] dump_stack+0x45/0x56
>>>> [ 2334.137734] [<ffffffff8106c93d>] warn_slowpath_common+0x7d/0xa0
>>>> [ 2334.138433] [<ffffffff8106ca1a>] warn_slowpath_null+0x1a/0x20
>>>> [ 2334.139109] [<ffffffffa0372103>] start_dma+0x153/0x160 [net2280]
>>>> [ 2334.139795] [<ffffffffa0373b1b>] net2280_queue+0x2db/0x480 [net2280]
>>>> [ 2334.140449] [<ffffffffa03b94f1>] start_transfer.isra.32+0x71/0xe0 [usb_f_mass_storage]
>>>> [ 2334.141128] [<ffffffffa03b959e>] start_out_transfer+0x3e/0x80 [usb_f_mass_storage]
>>>> [ 2334.141772] [<ffffffffa03ba967>] fsg_main_thread+0x207/0x17f0 [usb_f_mass_storage]
>>>> [ 2334.142374] [<ffffffff81749eda>] ? __schedule+0x37a/0x830
>>>> [ 2334.142950] [<ffffffffa03ba760>] ? handle_exception+0x460/0x460 [usb_f_mass_storage]
>>>> [ 2334.143534] [<ffffffff8108a3e2>] kthread+0xd2/0xf0
>>>> [ 2334.144054] [<ffffffff8108a310>] ? kthread_create_on_node+0x180/0x180
>>>> [ 2334.144595] [<ffffffff8174ecbc>] ret_from_fork+0x7c/0xb0
>>>> [ 2334.145155] [<ffffffff8108a310>] ? kthread_create_on_node+0x180/0x180
>>>> [ 2334.145699] ---[ end trace 4ded8927b95f4eec ]---
>>>> followed by:
>>>> [ 2396.154501] net2280 0000:01:00.0: The dmastat return = 5002!!
>>>> [ 2396.253653] g_mass_storage gadget: super-speed config #1: Linux File-Backed Storage
>>>>
>>>> Unplugging and re-plugging the USB cable does allow me to remount the volume.
>>>>
>>>> Paul.
>>>> On 25 Nov 2014, at 15:24, Ricardo Ribalda Delgado <ricardo.ribalda@xxxxxxxxx> wrote:
>>>>
>>>>> Could you try
>>>>> if (likely(t & (BIT(FIFO_EMPTY) | BIT(NAK_OUT_PACKETS)))){
>>>>>
>>>>> instead of
>>>>>
>>>>> if (likely(t & BIT(FIFO_EMPTY)))
>>>>>
>>>>>
>>>>> And tell if it works better
>>>>
>>>
>>>
>>>
>>> --
>>> Ricardo Ribalda
>>> --
>>> To unsubscribe from this list: send the line "unsubscribe linux-usb" in
>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
