* Re: [PATCH] ath10k: disable wake_tx_queue for older devices
From: Roman Yeryomin @ 2016-08-04 10:07 UTC (permalink / raw)
To: Dave Taht
Cc: Valo, Kalle, linux-wireless@vger.kernel.org,
michal.kazior@tieto.com, ath10k@lists.infradead.org
In-Reply-To: <CAA93jw6j0pBKnm3Fs=D7V1CUsBnhZKKymbk4=UQuFWKuEAs-+Q@mail.gmail.com>
On 1 August 2016 at 12:04, Dave Taht <dave.taht@gmail.com> wrote:
> On Mon, Aug 1, 2016 at 1:35 AM, Roman Yeryomin <leroi.lists@gmail.com> wrote:
>> On 7 July 2016 at 19:30, Valo, Kalle <kvalo@qca.qualcomm.com> wrote:
>>> Michal Kazior <michal.kazior@tieto.com> writes:
>>>
>>>> Ideally wake_tx_queue should be used regardless as
>>>> it is a requirement for reducing bufferbloat and
>>>> implementing airtime fairness in the future.
>>>>
>>>> However some setups (typically low-end platforms
>>>> hosting QCA988X) suffer performance regressions
>>>> with the current wake_tx_queue implementation.
>>>> Therefore disable it unless it is really
>>>> beneficial with current codebase (which is when
>>>> firmware supports smart pull-push tx scheduling).
>>>>
>>>> Signed-off-by: Michal Kazior <michal.kazior@tieto.com>
>>>
>>> I think it's too late to send this to 4.7 anymore (and this due to my
>>> vacation). So I'm planning to queue this to 4.8, but if the feedback is
>>> positive we can always send this to a 4.7 stable release.
>>>
>>
>> Sorry guys, drowned.
>> So, yes, applying this patch does the job. That is gets me to the
>> results similar to
>> https://lists.openwrt.org/pipermail/openwrt-devel/2016-May/041448.html
>>
>> Going to try latest code on same system...
>
> Can you try increasing the quantum to 1514, and reducing the codel
> target to 5ms? (without this patch?)
>
So it was 1514 already...
Regards,
Roman
^ permalink raw reply
* [PATCH] ath9k: fix AR5416 access GPIO warning
From: miaoqing @ 2016-08-04 7:48 UTC (permalink / raw)
To: kvalo; +Cc: linux-wireless, ath9k-devel, s.l-h, Miaoqing Pan
From: Miaoqing Pan <miaoqing@codeaurora.org>
The warning was seen on AR5416 chip, which invoke ath9k_hw_gio_get()
before the GPIO initialized correctly.
WARNING: CPU: 1 PID: 1159 at ~/drivers/net/wireless/ath/ath9k/hw.c:2776 ath9k_hw_gpio_get+0x148/0x1a0 [ath9k_hw]
...
CPU: 1 PID: 1159 Comm: systemd-udevd Not tainted 4.7.0-rc7-aptosid-amd64 #1 aptosid 4.7~rc7-1~git92.slh.3
Hardware name: /DH67CL, BIOS BLH6710H.86A.0160.2012.1204.1156 12/04/2012
0000000000000286 00000000f912d633 ffffffff81290fd3 0000000000000000
0000000000000000 ffffffff81063fd4 ffff88040c6dc018 0000000000000000
0000000000000002 0000000000000000 0000000000000100 ffff88040c6dc018
Call Trace:
[<ffffffff81290fd3>] ? dump_stack+0x5c/0x79
[<ffffffff81063fd4>] ? __warn+0xb4/0xd0
[<ffffffffa0668fb8>] ? ath9k_hw_gpio_get+0x148/0x1a0 [ath9k_hw]
Signed-off-by: Miaoqing Pan <miaoqing@codeaurora.org>
---
drivers/net/wireless/ath/ath9k/hw.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/net/wireless/ath/ath9k/hw.c b/drivers/net/wireless/ath/ath9k/hw.c
index d1d0c06..14b13f0 100644
--- a/drivers/net/wireless/ath/ath9k/hw.c
+++ b/drivers/net/wireless/ath/ath9k/hw.c
@@ -2482,6 +2482,8 @@ int ath9k_hw_fill_cap_info(struct ath_hw *ah)
return -EINVAL;
}
+ ath9k_gpio_cap_init(ah);
+
if (AR_SREV_9485(ah) ||
AR_SREV_9285(ah) ||
AR_SREV_9330(ah) ||
@@ -2531,8 +2533,6 @@ int ath9k_hw_fill_cap_info(struct ath_hw *ah)
else
pCap->hw_caps &= ~ATH9K_HW_CAP_HT;
- ath9k_gpio_cap_init(ah);
-
if (AR_SREV_9160_10_OR_LATER(ah) || AR_SREV_9100(ah))
pCap->rts_aggr_limit = ATH_AMPDU_LIMIT_MAX;
else
--
1.9.1
^ permalink raw reply related
* Re: Buggy rhashtable walking
From: Herbert Xu @ 2016-08-04 7:45 UTC (permalink / raw)
To: David S. Miller, netdev, Johannes Berg, linux-wireless; +Cc: Thomas Graf, tom
In-Reply-To: <20160804071846.GA773@gondor.apana.org.au>
On Thu, Aug 04, 2016 at 03:18:46PM +0800, Herbert Xu wrote:
>
> So the question is can wireless handle seeing an entry multiple
> times? In particular, __ieee80211_rx_handle_packet would appear
> to process the same packet multiple times if this were to happen.
It's worse than I thought. In fact it's not walking the table
at all, rather it's doing a hash lookup by hand!
This cannot possibly work given that rhashtable makes use of
multiple hash tables.
In fact this also demonstrates why putting multiple identical
objects into the same table is crap. Because there is no sane
way of returning all objects corresponding to a single key, given
that they may be spread over multiple tables.
So I'm going to fix this by consolidating identical objects into
a single rhashtable entry which also lets us get rid of the
insecure_elasticity setting.
So the next time someone comes along and wants to add multiple
objects with the same key to one table, please just say no.
Thanks,
--
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
^ permalink raw reply
* Buggy rhashtable walking
From: Herbert Xu @ 2016-08-04 7:18 UTC (permalink / raw)
To: David S. Miller, netdev, Johannes Berg, linux-wireless
Hi:
While working on rhashtable I noticed that wireless is walking
rhashtables by hand using rht_for_each_*. You must not do that
as an rhashtable can entail multiple hash tables when resizing.
If you walk it by hand then you may end up missing entries.
The correct way to do it is to use the rhashtable walk interface.
However, even this comes with the caveat that a given entry may
show up multiple times. So if you cannot handle that then you
must construct your own data structure outside of rhashtable, like
we do in IPsec.
So the question is can wireless handle seeing an entry multiple
times? In particular, __ieee80211_rx_handle_packet would appear
to process the same packet multiple times if this were to happen.
Cheers,
--
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
^ permalink raw reply
* Re: [PATCH v2] nl80211: Receive correct value for NL80211_MESHCONF_HT_OPMODE command
From: Johannes Berg @ 2016-08-04 5:50 UTC (permalink / raw)
To: Masashi Honma; +Cc: linux-wireless, j, me
In-Reply-To: <d193404d-8a77-8a8e-10eb-f46bd4008384@gmail.com>
> And I think checking only in wpa_supplicant is not good idea. Because
> other user application can access to the kernel API. If invalid flag
> combination causes kernel panic, it could be kernel vulnerability.
>
I don't really see how that should confuse a driver into a panic, but
fair enough.
johannes
^ permalink raw reply
* Re: [PATCH] ath9k: fix GPIO mask for AR9462 and AR9565
From: Stefan Lippers-Hollmann @ 2016-08-04 6:24 UTC (permalink / raw)
To: Pan, Miaoqing
Cc: Valo, Kalle, miaoqing@codeaurora.org,
linux-wireless@vger.kernel.org, ath9k-devel,
sudipm.mukherjee@gmail.com
In-Reply-To: <1470288583349.11756@qti.qualcomm.com>
Hi
On 2016-08-04, Pan, Miaoqing wrote:
> Sorry, try the patch. AR5416 will invoke ath9k_hw_gpio_get() before gpio initialized correctly.
>
> Thanks,
> Miaoqing
Thanks a lot, the patch appears to work. Feel free to add
Tested-by: Stefan Lippers-Hollmann <s.l-h@gmx.de>
and please also send it to stable-4.7 (confirmed working).
Cc: stable-4.7+ <stable@vger.kernel.org> #4.7+
Regards
Stefan Lippers-Hollmann
^ permalink raw reply
* Re: [PATCH] ath6kl: enable firmware crash dumps on the AR6004
From: Steve deRosier @ 2016-08-03 23:18 UTC (permalink / raw)
To: engineering; +Cc: Kalle Valo, ath6kl, linux-wireless
In-Reply-To: <1470257023-6747-1-git-send-email-dan.kephart@lairdtech.com>
On Wed, Aug 3, 2016 at 1:43 PM, <engineering@keppy.com> wrote:
> From: Dan Kephart <dan.kephart@lairdtech.com>
>
> The firmware crash dumps on the 6004 are the same as the 6003. Remove the
> statement guarding it from dumping on the 6004. Renamed the
> REG_DUMP_COUNT_AR6003 to reflect support on both chips.
>
> Signed-off-by: Dan Kephart <dan.kephart@lairdtech.com>
Since ath6kl only supports the 6003 and 6004, this looks fine to me
and works OK on my platforms.
Reviewed-by: Steve deRosier <steve.derosier@lairdtech.com>
^ permalink raw reply
* Re: [PATCH] ath9k: fix GPIO mask for AR9462 and AR9565
From: Pan, Miaoqing @ 2016-08-04 5:30 UTC (permalink / raw)
To: Valo, Kalle, Stefan Lippers-Hollmann, miaoqing@codeaurora.org
Cc: linux-wireless@vger.kernel.org, ath9k-devel,
sudipm.mukherjee@gmail.com
In-Reply-To: <87r3a6556p.fsf@qca.qualcomm.com>
[-- Attachment #1: Type: text/plain, Size: 2282 bytes --]
Sorry, try the patch. AR5416 will invoke ath9k_hw_gpio_get() before gpio initialized correctly.
Thanks,
Miaoqing
________________________________________
From: Valo, Kalle
Sent: Wednesday, August 3, 2016 9:54 PM
To: Stefan Lippers-Hollmann; miaoqing@codeaurora.org
Cc: linux-wireless@vger.kernel.org; ath9k-devel; sudipm.mukherjee@gmail.com
Subject: Re: [PATCH] ath9k: fix GPIO mask for AR9462 and AR9565
Stefan Lippers-Hollmann <s.l-h@gmx.de> writes:
> Hi
>
> On 2016-06-03, miaoqing@codeaurora.org wrote:
>> From: Miaoqing Pan <miaoqing@codeaurora.org>
>>
>> The incorrect GPIO mask cause kernel warning, when AR9462 access GPIO11.
>> Also fix the mask for AR9565.
> [...]
>
> I think I'm seeing a very similar issue on AR5008/ AR5416+AR2133 and
> 4.7-rc7 (mainline v4.7-rc7-92-g47ef4ad, to be exact).
>
> [ 4.958874] ath9k 0000:02:02.0: enabling device (0000 -> 0002)
> [...]
> [ 5.401086] ------------[ cut here ]------------
> [ 5.401093] WARNING: CPU: 1 PID: 1159 at /build/linux-aptosid-4.7~rc7/drivers/net/wireless/ath/ath9k/hw.c:2776 ath9k_hw_gpio_get+0x148/0x1a0 [ath9k_hw]
[...]
> Reverting this, and the other patches depending on it, fixes the
> problem for me:
>
> e024111f6946f45cf1559a8c6fd48d2d0f696d07 Revert "ath9k: fix GPIO mask for AR9462 and AR9565"
> db2221901fbded787daed153281ed875de489692 Revert "ath9k: free GPIO resource for SOC GPIOs"
> c7212b7136ba69efb9785df68b669381cb893920 Revert "ath9k: fix BTCoex configuration for SOC chips"
> dfcf02cd2998e2240b2bc7b4f4412578b8070bdb Revert "ath9k: fix BTCoex access invalid registers for SOC chips"
> 668ae0a3e48ac6811f431915b466514bf167e2f4 Revert "ath9k: add bits definition of BTCoex MODE2/3 for SOC chips"
> c8770bcf5cefa8cbfae21c07c4fe3428f5a9d42a Revert "ath9k: Allow platform override BTCoex pin"
> 79d4db1214a0c7b1818aaf64d0606b17ff1acea7 Revert "ath9k: cleanup led_pin initial"
> b2d70d4944c1789bc64376ad97a811f37e230c87 Revert "ath9k: make GPIO API to support both of WMAC and SOC"
> a01ab81b09c55025365c1de1345b941a18e05529 Revert "ath9k: define correct GPIO numbers and bits mask"
>
> AR9285 (168c:002b) is fine either way.
Miaoqing, have you looked at this? Looks like another regression which
should be fixed.
--
Kalle Valo
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: ath9k-fix-accessing-gpio-warning.patch --]
[-- Type: text/x-diff; name="ath9k-fix-accessing-gpio-warning.patch", Size: 538 bytes --]
--- a/drivers/net/wireless/ath/ath9k/hw.c
+++ b/drivers/net/wireless/ath/ath9k/hw.c
@@ -2482,6 +2482,8 @@ int ath9k_hw_fill_cap_info(struct ath_hw *ah)
return -EINVAL;
}
+ ath9k_gpio_cap_init(ah);
+
if (AR_SREV_9485(ah) ||
AR_SREV_9285(ah) ||
AR_SREV_9330(ah) ||
@@ -2531,8 +2533,6 @@ int ath9k_hw_fill_cap_info(struct ath_hw *ah)
else
pCap->hw_caps &= ~ATH9K_HW_CAP_HT;
- ath9k_gpio_cap_init(ah);
-
if (AR_SREV_9160_10_OR_LATER(ah) || AR_SREV_9100(ah))
pCap->rts_aggr_limit = ATH_AMPDU_LIMIT_MAX;
else
^ permalink raw reply
* Re: [PATCH] wl3501_cs: Add spinlock to wl3501_reset
From: Vaishali Thakkar @ 2016-08-04 5:01 UTC (permalink / raw)
To: Pavel Andrianov, Arnaldo Carvalho de Melo
Cc: Kalle Valo, linux-wireless, netdev, linux-kernel, ldv-project
In-Reply-To: <1470130913-5218-1-git-send-email-andrianov@ispras.ru>
On Tuesday 02 August 2016 03:11 PM, Pavel Andrianov wrote:
> Likely wl3501_reset should acquire spinlock as wl3501_{open, close}.
> One of calls of wl3501_reset has been already protected.
> The others were unprotected and might lead to a race condition.
> The patch adds spinlock into the wl3501_reset and removes it from
> wl3501_tx_timeout.
>
> Found by Linux Driver Verification project (linuxtesting.org)
>
> Signed-off-by: Pavel Andrianov <andrianov@ispras.ru>
Acked-by: Vaishali Thakkar <vaishali.thakkar@oracle.com>
> ---
> drivers/net/wireless/wl3501_cs.c | 7 +++----
> 1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/net/wireless/wl3501_cs.c b/drivers/net/wireless/wl3501_cs.c
> index 13fd734..196f13c 100644
> --- a/drivers/net/wireless/wl3501_cs.c
> +++ b/drivers/net/wireless/wl3501_cs.c
> @@ -1247,7 +1247,9 @@ static int wl3501_reset(struct net_device *dev)
> {
> struct wl3501_card *this = netdev_priv(dev);
> int rc = -ENODEV;
> + unsigned long flags;
>
> + spin_lock_irqsave(&this->lock, flags);
> wl3501_block_interrupt(this);
>
> if (wl3501_init_firmware(this)) {
> @@ -1269,20 +1271,17 @@ static int wl3501_reset(struct net_device *dev)
> pr_debug("%s: device reset", dev->name);
> rc = 0;
> out:
> + spin_unlock_irqrestore(&this->lock, flags);
> return rc;
> }
>
> static void wl3501_tx_timeout(struct net_device *dev)
> {
> - struct wl3501_card *this = netdev_priv(dev);
> struct net_device_stats *stats = &dev->stats;
> - unsigned long flags;
> int rc;
>
> stats->tx_errors++;
> - spin_lock_irqsave(&this->lock, flags);
> rc = wl3501_reset(dev);
> - spin_unlock_irqrestore(&this->lock, flags);
> if (rc)
> printk(KERN_ERR "%s: Error %d resetting card on Tx timeout!\n",
> dev->name, rc);
>
--
Vaishali
^ permalink raw reply
* Re: [PATCH v2] nl80211: Receive correct value for NL80211_MESHCONF_HT_OPMODE command
From: Masashi Honma @ 2016-08-04 0:37 UTC (permalink / raw)
To: Johannes Berg; +Cc: linux-wireless, j, me
In-Reply-To: <1470207156.2638.9.camel@sipsolutions.net>
On 2016年08月03日 15:52, Johannes Berg wrote:
> I'm actually half thinking that we could just remove all restrictions
> on this and allow any u16 value of this field, and rely on
> wpa_supplicant to do the right thing... Then we don't have to update
> this if we ever want to do something new either.
>
> What do you think? What does the validation actually help us with?
I think checking the bits here is better than allowing all values.
Because if we allow any values for ht_opmode, kernel developer needs to
care about any bit combination working well. For example, kernel
developer should test there is not any unexpected thing when non-GF and
non-HT both flags are enabled. If we check invalid bit at the entrance,
we don't need to care anymore about invalid combination. In any case we
need to care about combination. Then, it is more easy to do it near the
entrance.
And I think checking only in wpa_supplicant is not good idea. Because
other user application can access to the kernel API. If invalid flag
combination causes kernel panic, it could be kernel vulnerability.
Masashi Honma.
^ permalink raw reply
* Re: PROBLEM: network data corruption (bisected to e5a4b0bb803b)
From: Alan Curry @ 2016-08-03 23:25 UTC (permalink / raw)
To: Christian Lamparter
Cc: Alan Curry, Al Viro, alexmcwhirter, David Miller, linux-wireless,
netdev, linux-kernel
In-Reply-To: <1882749.mb7lsAROoU@debian64>
Christian Lamparter wrote:
>
> One thing I noticed in your previous post is that you "might" not have
> draft-802.11n enabled. Do you see any "disabling HT/VHT due to WEP/TKIP use."
> in your dmesg logs? If so, check if you can force your AP to use WPA2
> with CCMP/AES only.
>
Yes, I've had that message. The reason wan't on the AP though. My
wpa_supplicant.conf only had TKIP enabled, because that's what was in the
sample configuration file I started with. Adding CCMP there worked, and in
that mode I'm no longer getting any corrupted packets.
If I'd paid attention to the encryption options when setting up this network
originally, I would have had CCMP the whole time, with no corrupted packets,
and never would have found the iov iterator bug...
--
Alan Curry
^ permalink raw reply
* [PATCH] ath6kl: enable firmware crash dumps on the AR6004
From: engineering @ 2016-08-03 20:43 UTC (permalink / raw)
To: kvalo; +Cc: ath6kl, linux-wireless
From: Dan Kephart <dan.kephart@lairdtech.com>
The firmware crash dumps on the 6004 are the same as the 6003. Remove the
statement guarding it from dumping on the 6004. Renamed the
REG_DUMP_COUNT_AR6003 to reflect support on both chips.
Signed-off-by: Dan Kephart <dan.kephart@lairdtech.com>
---
drivers/net/wireless/ath/ath6kl/hif.c | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)
diff --git a/drivers/net/wireless/ath/ath6kl/hif.c b/drivers/net/wireless/ath/ath6kl/hif.c
index 18c0708..d194253 100644
--- a/drivers/net/wireless/ath/ath6kl/hif.c
+++ b/drivers/net/wireless/ath/ath6kl/hif.c
@@ -64,7 +64,7 @@ int ath6kl_hif_rw_comp_handler(void *context, int status)
}
EXPORT_SYMBOL(ath6kl_hif_rw_comp_handler);
-#define REG_DUMP_COUNT_AR6003 60
+#define REGISTER_DUMP_COUNT 60
#define REGISTER_DUMP_LEN_MAX 60
static void ath6kl_hif_dump_fw_crash(struct ath6kl *ar)
@@ -73,9 +73,6 @@ static void ath6kl_hif_dump_fw_crash(struct ath6kl *ar)
u32 i, address, regdump_addr = 0;
int ret;
- if (ar->target_type != TARGET_TYPE_AR6003)
- return;
-
/* the reg dump pointer is copied to the host interest area */
address = ath6kl_get_hi_item_addr(ar, HI_ITEM(hi_failure_state));
address = TARG_VTOP(ar->target_type, address);
@@ -95,7 +92,7 @@ static void ath6kl_hif_dump_fw_crash(struct ath6kl *ar)
/* fetch register dump data */
ret = ath6kl_diag_read(ar, regdump_addr, (u8 *)®dump_val[0],
- REG_DUMP_COUNT_AR6003 * (sizeof(u32)));
+ REGISTER_DUMP_COUNT * (sizeof(u32)));
if (ret) {
ath6kl_warn("failed to get register dump: %d\n", ret);
return;
@@ -105,9 +102,9 @@ static void ath6kl_hif_dump_fw_crash(struct ath6kl *ar)
ath6kl_info("hw 0x%x fw %s\n", ar->wiphy->hw_version,
ar->wiphy->fw_version);
- BUILD_BUG_ON(REG_DUMP_COUNT_AR6003 % 4);
+ BUILD_BUG_ON(REGISTER_DUMP_COUNT % 4);
- for (i = 0; i < REG_DUMP_COUNT_AR6003; i += 4) {
+ for (i = 0; i < REGISTER_DUMP_COUNT; i += 4) {
ath6kl_info("%d: 0x%8.8x 0x%8.8x 0x%8.8x 0x%8.8x\n",
i,
le32_to_cpu(regdump_val[i]),
--
2.4.10
^ permalink raw reply related
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Bjorn Andersson @ 2016-08-03 22:26 UTC (permalink / raw)
To: Luis R. Rodriguez
Cc: Daniel Wagner, Andrew Morton, Jeff Mahoney, Dmitry Torokhov,
Arend van Spriel, Daniel Wagner, Bastien Nocera,
Greg Kroah-Hartman, Johannes Berg, Kalle Valo, Ohad Ben-Cohen,
Mimi Zohar, David Howells, Andy Lutomirski, David Woodhouse,
Julia Lawall, linux-input, linux-kselftest, linux-wireless,
linux-kernel
In-Reply-To: <20160803155540.GL3296@wotan.suse.de>
On Wed 03 Aug 08:55 PDT 2016, Luis R. Rodriguez wrote:
> On Wed, Aug 03, 2016 at 08:57:09AM +0200, Daniel Wagner wrote:
> > On 08/02/2016 09:41 AM, Luis R. Rodriguez wrote:
[..]
> > Not sure if I get you here correctly. Is the 'system configurable
> > deterministic file' is a knob which controlled by user space? Or it
> > this something you define at compile time?
>
> I meant at compile time on the kernel. So CONFIG_READ_READY_SENTINEL
> or something like this, and it be a string, which if set then when
> the kernel read APIs are used, then a new API could be introduced
> that would *only* enable reading through once that sentinel has
> been detected by the kernel to allowed through reads. Doing this
> per mount / target filesystem is rather cumbersome given possible
> overlaps in mounts and also pivot_root() being possible, so instead
> targeting simply the fs/exec.c enum kernel_read_file_id would seem
> more efficient and clean but we would need a decided upon set of
> paths per enum kernel_read_file_id as base (or just one path per
> enum kernel_read_file_id). For number of paths I mean the number
> of target directories to look for the sentinel per enum kernel_read_file_id,
> so for instance for READING_FIRMWARE perhaps just deciding on /lib/firmware/
> would suffice, but if this supported multiple paths another option may be
> for the sentinel to also be looked for in /lib/firmware/updates/,
> /lib/firmware/" UTS_RELEASE -- etc. It would *stop* after finding one
> sentinel on any of these paths.
>
> If a system has has CONFIG_READ_READY_SENTINEL it would mean an agreed upon
> system configuration has been decided so that at any point in time reads
> against READING_FIRMWARE using a new kernel_read_file_from_path_sentinel()
> (or something like it) would only allow the read to go through once
> the sentinel has been found for READING_FIRMWARE on the agreed upon
> paths.
>
> The benefit of the sentintel approach is it avoids complexities with
> pivot_root(), and makes the deterministic aspect of the target left
> only to a system-configuration enabled target path / file.
>
This sounds reasonable, it could be configured to wait for a certain
static file or userspace could generate this file once it reaches some
checkpoint.
Just to provide some additional input to "will rootfs mounted be enough
of a sentinel".
In an Android device you have a initramfs that will read an fstab file
and mount /system that holds most firmware, for some platforms
additional firmware will come from a third partition (in the Qualcomm
case mounted in /persist by the same mechanism).
With the sentinel approach one could configure the system either point
it at a file in the last file system to be mounted or have init generate
a file once its done with this; or in the generic configuration just
wait for /lib/firmware to show up.
I like this approach.
> This is just an idea. I'd like some FS folks to review.
>
Regards,
Bjorn
^ permalink raw reply
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Arend van Spriel @ 2016-08-03 20:42 UTC (permalink / raw)
To: Dmitry Torokhov
Cc: Luis R. Rodriguez, Daniel Wagner, Bjorn Andersson, Daniel Wagner,
Bastien Nocera, Greg Kroah-Hartman, Johannes Berg, Kalle Valo,
Ohad Ben-Cohen, Mimi Zohar, David Howells, Andy Lutomirski,
David Woodhouse, Julia Lawall, linux-input@vger.kernel.org,
linux-kselftest, linux-wireless, lkml
In-Reply-To: <20160803153506.GC29702@dtor-ws>
On 03-08-16 17:35, Dmitry Torokhov wrote:
>> In my opinion the kernel should provide functionality to user-space and
>> > user-space providing functionality to the kernel should be avoided.
> Why? We have bunch of stuff running in userspace for the kernel. Fuse
> for example. I am sure there are more.
To me "running in user-space" is not the same as providing
functionality, but I see your point given below.
>> >
>>> > > If we solve waiting for rootfs (or something else that may contain
>>> > > firmware) then these cases will not need to use usermode helper.
>> >
>> > If firmware (or whatever) API could get notification of mount syscall it
>> > could be used to retry firmware loading instead of periodic polling.
>> > That leaves the question raised by you about when to stop trying. The
>> > initlevel stuff is probably a user-space only concept, right? So no
>> > ideas how the kernel itself could decide except for a "long" timeout.
> The kernel really does not know, it can only guess. The firmware may get
> delivered by motorized carrier pidgeons. But distribution does know how
> they set up, so they are in position to tell the kernel "go" or "give
> up".
What distro employs pidgeons. Like to give it a spin ;-)
Maybe the latest idea from Luis is a viable option.
Regards,
Arend
^ permalink raw reply
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Luis R. Rodriguez @ 2016-08-03 18:08 UTC (permalink / raw)
To: Bjorn Andersson, Jeff Mahoney, Takashi Iwai, Andrew Morton
Cc: Luis R. Rodriguez, Daniel Wagner, Dmitry Torokhov,
Arend van Spriel, Daniel Wagner, Bastien Nocera,
Greg Kroah-Hartman, Johannes Berg, Kalle Valo, Ohad Ben-Cohen,
Mimi Zohar, David Howells, Andy Lutomirski, David Woodhouse,
Julia Lawall, linux-input, linux-kselftest, linux-wireless,
linux-kernel
In-Reply-To: <20160803173955.GD13516@tuxbot>
On Wed, Aug 03, 2016 at 10:39:55AM -0700, Bjorn Andersson wrote:
> On Tue 02 Aug 00:41 PDT 2016, Luis R. Rodriguez wrote:
>
> > On Tue, Aug 02, 2016 at 08:53:55AM +0200, Daniel Wagner wrote:
> > > On 08/02/2016 08:34 AM, Luis R. Rodriguez wrote:
> > > >On Tue, Aug 02, 2016 at 07:49:19AM +0200, Daniel Wagner wrote:
> > > >>>The sysdata API's main goal rather is to provide a flexible API first,
> > > >>>compartamentalizing the usermode helper was secondary. But now it seems
> > > >>>I may just also add devm support too to help simplify code further.
> > > >>
> > > >>I missed the point that you plan to add usermode helper support to
> > > >>the sysdata API.
> > > >
> > > >I had no such plans, when I have asked folks so far about "hey are you
> > > >really in need for it, OK what for? " and "what extended uses do you
> > > >envision?" so I far I have not gotten any replies at all. So -- instead
> > > >sysdata currently ignores it.
> > >
> > > So you argue for the remoteproc use case with 100+ MB firmware that
> > > if there is a way to load after pivot_root() (or other additional
> > > firmware partition shows up) then there is no need at all for
> > > usermode helper?
> >
> > No, I'm saying I'd like to hear valid uses cases for the usermode helper and so
> > far I have only found using coccinelle grammar 2 explicit users, that's it. My
> > patch series (not yet merge) then annotates these as valid as I've verified
> > through their documentation they have some quirky requirement.
> >
>
> I think we're on the same page, but just to make sure; I do not want the
> usermode helper,
Yay.
> I only want a way to wait for the firmware files to
> become available.
Sure.
> > Other than these two drivers I'd like hear to valid requirements for it.
> >
> > The existential issue is a real issue but it does not look impossible to
> > resolve. It may be a solution to bloat up the kernel with 100+ MB size just to
> > stuff built-in firmware to avoid this issue, but it does not mean a solution
> > is not possible.
> >
> > Remind me -- why can remoteproc not stuff the firmware in initramfs ?
> >
>
> RAM usage:
> Storing the files in initramfs would consume 100MB RAM, we would then
> allocate 100MB RAM for buffers during firmware loading and then we have
> the reserved 100MB for the peripherals. The buffers could be easily be
> removed with a mechanism for providing a buffer to the load operation,
> but we would still double the RAM consumption.
>
> Boot time:
> Enlarging the kernel by 100MB will give noticeable addition to boot
> times.
Right I see.. Since we read the full kernel...
> Development issues:
> I have numerous concerns related to this, e.g. not being able to side
> load the firmware files without rebuilding the initramfs. But most of
> these are not technical issues, but rather a matter of convenience.
>
> One large issue would be how to figure out how large to make the boot
> partition in your Android phone, to cope with potential future growth in
> firmware size - which has already proven to be a mess.
>
> Legal matters:
> Some of these firmware files are not redistributable, making it
> impossible for end users to rebuild their kernel without loosing
> functionality. There are even cases where these files are not allowed to
> share partition with GPL binaries.
>
>
> Most of these are just a major inconveniences to the developer but some
> are show stoppers; especially the legal matters. So if we wave this off
> as something people can live without then every downstream will sit on
> their own solution to reimplement it.
Thanks I'll document these into the firmware_class.
> > Anyway, here's a simple suggestion: fs/exec.c gets a sentinel file monitor
> > support per enum kernel_read_file_id. For instance we'd have one for
> > READING_FIRMWARE, one for READING_KEXEC_IMAGE, perhaps READING_POLICY, and this
> > would in turn be used as the system configurable deterministic file for
> > which to wait for to be present before enabling each enum kernel_read_file_id
> > type read.
> >
> > Thoughts ?
>
> That does sound like a good generic solution for our problem and for the
> other types of files as well. Do you have any ideas (patches?) on how
> each sentinel would be triggered?
>
> The only concern I can think of right now is that the
> firmware_class.path might point to a separate partition; but based on
> how the signaling of the sentinels are implemented this might not be an
> issue.
There's another simpler suggestion I'm getting too, will post in the other
thread.
Luis
^ permalink raw reply
* Re: [RFC] ath10k: silence firmware file probing warnings
From: Arend van Spriel @ 2016-08-03 19:19 UTC (permalink / raw)
To: Luis R. Rodriguez, Valo, Kalle
Cc: Stanislaw Gruszka, Prarit Bhargava, Greg Kroah-Hartman, Ming Lei,
linux-wireless, ath10k, mmarek@suse.com, michal.kazior@tieto.com,
Arend van Spriel, Emmanuel Grumbach
In-Reply-To: <20160803171044.GP3296@wotan.suse.de>
On 03-08-16 19:10, Luis R. Rodriguez wrote:
> On Wed, Aug 03, 2016 at 03:04:39PM +0000, Valo, Kalle wrote:
>> "Luis R. Rodriguez" <mcgrof@kernel.org> writes:
>>
>>> On Wed, Aug 03, 2016 at 01:33:31PM +0200, Arend van Spriel wrote:
>>>> On 02-08-16 16:16, Luis R. Rodriguez wrote:
>>>>> On Tue, Aug 02, 2016 at 11:10:22AM +0000, Valo, Kalle wrote:
>>>>>> "Luis R. Rodriguez" <mcgrof@kernel.org> writes:
>>>>>>
>>>>>>> I was considering this as a future extension to the firmware API
>>>>>>> through the new extensible firmware API, the sysdata API.
>>>>>>
>>>>>> I think Linus mentioned this already, but I want to reiterate anyway.
>>>>>> The name "sysdata" is horrible, I didn't have any idea what it means
>>>>>> until I read your description. Please continue to use the term
>>>>>> "firmware", anyone already know what it means.
>>>>>
>>>>> We've gone well past using the firmware API for firmware though, if
>>>>> we use it for 802.11 to replace CRDA for instance its really odd to
>>>>> be calling it firmware. But sure... I will rebrand again to firmware...
>>>>
>>>> I tend to agree. Although some people even call an OpenWrt image
>>>> firmware. Guess it is just in the eye of the beholder.
>>>
>>> Sure...
>>>
>>> Come to think of it I'll still go with "sysdata", this is a very minor
>>> detail, do let me know if there is anything technical rather than
>>> the color of the bikeshed [0] over the patches.
>>
>> Well, you don't seem to care but I prefer that the terminology is clear
>> and I don't want to waste people's time browsing the source to find out
>> what something means.
>
> Its not that I don't care, its this is a super trivial matter, like the
> color of a bikeshed, and I'd much prefer to put energy and review on
> technical matters.
>
>> Even "driverdata" would be more descriptive for me
>> than "sysdata".
>>
>> Actually, what does the "sys" refer here, system? And what system is
>> that exactly?
>
> Yes system, so as in system data file. "driver_data" is just as good.
> Although who knows, others may want to paint the bikeshed a different
> color.
>
> The core reason why the name change is to make emphasis of the fact that
> we've gone way past the point where the APIs are used for non-firmware
> and I expect this use will only grow.
Here some colors to add to the palet with "color code" in brackets:
- device specific data (dsd): although if CRDA stuff is loaded it is no
longer tied to a device.
- eXtensible firmware API (xfw): the ever popular x factor :-p
- binary blob loader (bbl): just popped up in me head.
Need more beer to go on.
Regards,
Arend
^ permalink raw reply
* [patch] ath9k: indent an if statement
From: Dan Carpenter @ 2016-08-03 18:43 UTC (permalink / raw)
To: QCA ath9k Development, Benjamin Berg
Cc: Kalle Valo, linux-wireless, ath9k-devel, kernel-janitors
It looks like this code is correct, but it just needs to be indented a
bit.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
diff --git a/drivers/net/wireless/ath/ath9k/main.c b/drivers/net/wireless/ath/ath9k/main.c
index a394622..eb00724 100644
--- a/drivers/net/wireless/ath/ath9k/main.c
+++ b/drivers/net/wireless/ath/ath9k/main.c
@@ -919,7 +919,7 @@ static void ath9k_vif_iter_set_beacon(struct ath9k_vif_iter_data *iter_data,
} else {
if (iter_data->primary_beacon_vif->type != NL80211_IFTYPE_AP &&
vif->type == NL80211_IFTYPE_AP)
- iter_data->primary_beacon_vif = vif;
+ iter_data->primary_beacon_vif = vif;
}
iter_data->beacons = true;
^ permalink raw reply related
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Luis R. Rodriguez @ 2016-08-03 18:40 UTC (permalink / raw)
To: Daniel Wagner, Bjorn Andersson, Tom Gundersen, Dmitry Torokhov,
Arend van Spriel, kay, Hannes Reinecke
Cc: Daniel Wagner, Andrew Morton, Jeff Mahoney, Daniel Wagner,
Bastien Nocera, Greg Kroah-Hartman, Johannes Berg, Kalle Valo,
Ohad Ben-Cohen, Mimi Zohar, David Howells, Andy Lutomirski,
David Woodhouse, Julia Lawall, linux-input, linux-kselftest,
linux-wireless, linux-kernel, Luis R. Rodriguez
In-Reply-To: <20160803155540.GL3296@wotan.suse.de>
On Wed, Aug 03, 2016 at 05:55:40PM +0200, Luis R. Rodriguez wrote:
> On Wed, Aug 03, 2016 at 08:57:09AM +0200, Daniel Wagner wrote:
> > On 08/02/2016 09:41 AM, Luis R. Rodriguez wrote:
> > >On Tue, Aug 02, 2016 at 08:53:55AM +0200, Daniel Wagner wrote:
> > >>On 08/02/2016 08:34 AM, Luis R. Rodriguez wrote:
> > >>>On Tue, Aug 02, 2016 at 07:49:19AM +0200, Daniel Wagner wrote:
> > >>So you argue for the remoteproc use case with 100+ MB firmware that
> > >>if there is a way to load after pivot_root() (or other additional
> > >>firmware partition shows up) then there is no need at all for
> > >>usermode helper?
> > >
> > >No, I'm saying I'd like to hear valid uses cases for the usermode helper and so
> > >far I have only found using coccinelle grammar 2 explicit users, that's it. My
> > >patch series (not yet merge) then annotates these as valid as I've verified
> > >through their documentation they have some quirky requirement.
> >
> > I got that question wrong. It should read something like 'for the
> > remoteproc 100+MB there is no need for the user help?'.
>
> That's not a question for me but for those who say that the usermode helper
> is needed for remoteproc, so far from what folks are saying it seems the only
> reason for the usermodehelper was to try to avoid the deterministic issue,
> but I suggested a way to resolve that without the usermode helper now so
> would be curious to hear if there are any more reasons for it.
>
> > I've gone
> > through your patches and they make perfectly sense too. Maybe I can
> > convince you to take a better version of my patch 3 into your queue.
> > And I help you converting the exiting drivers. Obviously if you like
> > my help at all.
>
> I accept all help and would be glad to make enhancements instead of
> the old API through new API. The biggest thing here first I think is
> adding devm support, that I think should address what seemed to be
> the need to add more code for a transformation into the API. I'd
> personally only want to add that and be done with an introduction
> of the sysdata API. Further changes IMHO are best done atomically
> after that on top of it, but I'm happy to queue in the changes.
>
> > >Other than these two drivers I'd like hear to valid requirements for it.
> > >
> > >The existential issue is a real issue but it does not look impossible to
> > >resolve. It may be a solution to bloat up the kernel with 100+ MB size just to
> > >stuff built-in firmware to avoid this issue, but it does not mean a solution
> > >is not possible.
> > >
> > >Remind me -- why can remoteproc not stuff the firmware in initramfs ?
> >
> > I don't know. I was just bringing it up with the hope that Bjorn
> > will defend it. It seems my tactics didn't work out :)
>
> OK.
>
> > >Anyway, here's a simple suggestion: fs/exec.c gets a sentinel file monitor
> > >support per enum kernel_read_file_id. For instance we'd have one for
> > >READING_FIRMWARE, one for READING_KEXEC_IMAGE, perhaps READING_POLICY, and this
> > >would in turn be used as the system configurable deterministic file for
> > >which to wait for to be present before enabling each enum kernel_read_file_id
> > >type read.
> > >
> > >Thoughts ?
> >
> > Not sure if I get you here correctly. Is the 'system configurable
> > deterministic file' is a knob which controlled by user space? Or it
> > this something you define at compile time?
>
> I meant at compile time on the kernel. So CONFIG_READ_READY_SENTINEL
> or something like this, and it be a string, which if set then when
> the kernel read APIs are used, then a new API could be introduced
> that would *only* enable reading through once that sentinel has
> been detected by the kernel to allowed through reads. Doing this
> per mount / target filesystem is rather cumbersome given possible
> overlaps in mounts and also pivot_root() being possible, so instead
> targeting simply the fs/exec.c enum kernel_read_file_id would seem
> more efficient and clean but we would need a decided upon set of
> paths per enum kernel_read_file_id as base (or just one path per
> enum kernel_read_file_id). For number of paths I mean the number
> of target directories to look for the sentinel per enum kernel_read_file_id,
> so for instance for READING_FIRMWARE perhaps just deciding on /lib/firmware/
> would suffice, but if this supported multiple paths another option may be
> for the sentinel to also be looked for in /lib/firmware/updates/,
> /lib/firmware/" UTS_RELEASE -- etc. It would *stop* after finding one
> sentinel on any of these paths.
>
> If a system has has CONFIG_READ_READY_SENTINEL it would mean an agreed upon
> system configuration has been decided so that at any point in time reads
> against READING_FIRMWARE using a new kernel_read_file_from_path_sentinel()
> (or something like it) would only allow the read to go through once
> the sentinel has been found for READING_FIRMWARE on the agreed upon
> paths.
>
> The benefit of the sentintel approach is it avoids complexities with
> pivot_root(), and makes the deterministic aspect of the target left
> only to a system-configuration enabled target path / file.
>
> This is just an idea. I'd like some FS folks to review.
>
> > Hmm, so it would allow to decided to ask a userspace helper or load
> > the firmware directly (to be more precised the kernel_read_file_id
> > type). If yes, than it is what currently already have just
> > integrated nicely into the new sysdata API.
>
> Sorry, no, the above description is better of what I meant. This
> actually would not need to go into the sysdata API, unless of
> course we wanted it just as a new "feature" of it, but I don't
> think that's needed unless it has some implications behind the
> scenes. Given that firmware_class now uses a common core kernel
> API for reading files kernel_read_file_from_path() we could
> for instance add kernel_read_file_from_sentintel() and only
> if CONFIG_READ_READY_SENTINEL() would it block and wait until
> the sentinel clears. This should mean being able to make the
> change for both the old API and the new proposed sysdata API.
> Likewise for other kernel_read_file*() users -- they'd benefit
> from it as well.
A file sentinel would implicate a file namespace thing being used on the
filesystem -- to me this just means the Linux distribution / system integrator
would add this per filesystem, but agree this is pretty hacky. Furthermore
we'd wait forever if the Linux distribution / system integrator forgot to set
the sentinel file. That's not good. To avoid that a generic "root fs ready"
event could be sent from userspace to know when to clear stale reads... but if
that's going to be done best just replace all sentintels with a simple "root fs
ready" which would mean all reads from the kernel are ready. If we wanted
further granularity I suppose we could further have one event per enum
kernel_read_file_id, and a generic all-is-ready one.
To start off with then a simple event from userspace should suffice. But do keep
in mind that granularity might help given that a big iron system might have some
large array of disks to mount during bootup and that may take a long while, and
you likely want to read /lib/firmware way before that filesystem is ready.
Not sure if granularity fixated by enum kernel_read_file_id should suffice,
perhaps given its also enough for LSMs...
This indeed would mean a kernelspace and userspace change, but it would mean
not having to deal with the usermode helper crap anymore.
Anyway -- these are just ideas, patches welcomed !
Luis
^ permalink raw reply
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Luis R. Rodriguez @ 2016-08-03 17:37 UTC (permalink / raw)
To: Dmitry Torokhov, Tejun Heo, Linus Torvalds
Cc: Luis R. Rodriguez, Daniel Wagner, Andrew Morton, Jeff Mahoney,
Arend van Spriel, Bjorn Andersson, Daniel Wagner, Bastien Nocera,
Greg Kroah-Hartman, Johannes Berg, Kalle Valo, Ohad Ben-Cohen,
Mimi Zohar, David Howells, Andy Lutomirski, David Woodhouse,
Julia Lawall, linux-input, linux-kselftest, linux-wireless,
linux-kernel
In-Reply-To: <20160803161821.GB32965@dtor-ws>
On Wed, Aug 03, 2016 at 09:18:21AM -0700, Dmitry Torokhov wrote:
> On Wed, Aug 03, 2016 at 05:55:40PM +0200, Luis R. Rodriguez wrote:
> >
> > I accept all help and would be glad to make enhancements instead of
> > the old API through new API. The biggest thing here first I think is
> > adding devm support, that I think should address what seemed to be
> > the need to add more code for a transformation into the API. I'd
>
> I am confused. Why do we need devm support, given that devm is only
> valid in probe() paths[*] and we do know that we do not want to load
> firmware in probe() paths because it may cause blocking?
Its a good point, I hadn't gone on to implement devm support on the sysdata API
yet here so this requirement was not known to me. This certainly would put a
limitation to the idea of using devm then to deal with the firmware for you,
given that not all users of firmware are on probe, and as you note we want to
by default avoid firmware calls on probe since init+probe are called serially
by default unless a driver is using the new async probe. Nevertheless, even if
we had userspace or the driver always asking for async probe, most users of the
firmware API are not on probe anyway, so the gains of using devm to help with
freeing the firmware for the driver on probe would be very limited.
With that in mind, in retrospect then the current sysdata approach to require a
callback for synchronous calls would seem to work around this issue and
generalize a solution given we'd have:
For the sync case:
const struct sysdata_file_desc sysdata_desc = {
SYSDATA_DEFAULT_SYNC(driver_sync_req_cb, dev),
.keep = false, /* not explicitly needed as default is false */
};
ret = sysdata_file_request();
...
Behind the scenes firmware_class would call driver_sync_req_cb(),
since that's where we know the firmware will be consumed and since
the driver has explicitly asked that it no longer needs to keep the
firmware around (keep == false), it will free it on behalf of the
driver.
Since current synchronous calls for firmware do not have a callback
this would mean a driver changing to the sysdata API if it wanted
to take advantage of this feature of letting firmware_class free
the firmware for you, you'd need a bit more code than before.
For the asynchronous case this is a bit different given that the
current async firmware API requires a callback, so if keep == false
on the async sysdata API we just remove the release_firmware()
calls when converting over.
Given this, other than the bikeshedding aspects [0] ("sysdata", "driver data",
"firmware), perhaps the sysdata API is done then.
[0] http://phk.freebsd.dk/sagas/bikeshed.html
> [*] Yes, I know there are calls to devm* outside of probe() but I am
> pretty sure they are buggy unless they explicitly freed with devm* as
> well and then there is no point.
Really ? If so that's good to know.. and it should mean grammar could
be used to hunt this down, specially since we have now some grammar
basics to help us check for calls on probe or init. On the grammar
we'd just only complain if a call was used not in a probe path.
> IN all other cases it is likely wrong
> as it messes up with order of freeing resources.
Good to know, thanks. Hopefully the above semantics of the driver
using keep should suffice. Which gets me to think, what if devm
had something similar to white-list uses outside of probe so that
if a keep (or another flag name) was set then its vetting that
the order of freeing of resources is understood and fine.
Luis
^ permalink raw reply
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Bjorn Andersson @ 2016-08-03 17:39 UTC (permalink / raw)
To: Luis R. Rodriguez
Cc: Daniel Wagner, Dmitry Torokhov, Arend van Spriel, Daniel Wagner,
Bastien Nocera, Greg Kroah-Hartman, Johannes Berg, Kalle Valo,
Ohad Ben-Cohen, Mimi Zohar, David Howells, Andy Lutomirski,
David Woodhouse, Julia Lawall, linux-input, linux-kselftest,
linux-wireless, linux-kernel
In-Reply-To: <20160802074106.GI3296@wotan.suse.de>
On Tue 02 Aug 00:41 PDT 2016, Luis R. Rodriguez wrote:
> On Tue, Aug 02, 2016 at 08:53:55AM +0200, Daniel Wagner wrote:
> > On 08/02/2016 08:34 AM, Luis R. Rodriguez wrote:
> > >On Tue, Aug 02, 2016 at 07:49:19AM +0200, Daniel Wagner wrote:
> > >>>The sysdata API's main goal rather is to provide a flexible API first,
> > >>>compartamentalizing the usermode helper was secondary. But now it seems
> > >>>I may just also add devm support too to help simplify code further.
> > >>
> > >>I missed the point that you plan to add usermode helper support to
> > >>the sysdata API.
> > >
> > >I had no such plans, when I have asked folks so far about "hey are you
> > >really in need for it, OK what for? " and "what extended uses do you
> > >envision?" so I far I have not gotten any replies at all. So -- instead
> > >sysdata currently ignores it.
> >
> > So you argue for the remoteproc use case with 100+ MB firmware that
> > if there is a way to load after pivot_root() (or other additional
> > firmware partition shows up) then there is no need at all for
> > usermode helper?
>
> No, I'm saying I'd like to hear valid uses cases for the usermode helper and so
> far I have only found using coccinelle grammar 2 explicit users, that's it. My
> patch series (not yet merge) then annotates these as valid as I've verified
> through their documentation they have some quirky requirement.
>
I think we're on the same page, but just to make sure; I do not want the
usermode helper, I only want a way to wait for the firmware files to
become available.
> Other than these two drivers I'd like hear to valid requirements for it.
>
> The existential issue is a real issue but it does not look impossible to
> resolve. It may be a solution to bloat up the kernel with 100+ MB size just to
> stuff built-in firmware to avoid this issue, but it does not mean a solution
> is not possible.
>
> Remind me -- why can remoteproc not stuff the firmware in initramfs ?
>
RAM usage:
Storing the files in initramfs would consume 100MB RAM, we would then
allocate 100MB RAM for buffers during firmware loading and then we have
the reserved 100MB for the peripherals. The buffers could be easily be
removed with a mechanism for providing a buffer to the load operation,
but we would still double the RAM consumption.
Boot time:
Enlarging the kernel by 100MB will give noticeable addition to boot
times.
Development issues:
I have numerous concerns related to this, e.g. not being able to side
load the firmware files without rebuilding the initramfs. But most of
these are not technical issues, but rather a matter of convenience.
One large issue would be how to figure out how large to make the boot
partition in your Android phone, to cope with potential future growth in
firmware size - which has already proven to be a mess.
Legal matters:
Some of these firmware files are not redistributable, making it
impossible for end users to rebuild their kernel without loosing
functionality. There are even cases where these files are not allowed to
share partition with GPL binaries.
Most of these are just a major inconveniences to the developer but some
are show stoppers; especially the legal matters. So if we wave this off
as something people can live without then every downstream will sit on
their own solution to reimplement it.
> Anyway, here's a simple suggestion: fs/exec.c gets a sentinel file monitor
> support per enum kernel_read_file_id. For instance we'd have one for
> READING_FIRMWARE, one for READING_KEXEC_IMAGE, perhaps READING_POLICY, and this
> would in turn be used as the system configurable deterministic file for
> which to wait for to be present before enabling each enum kernel_read_file_id
> type read.
>
> Thoughts ?
That does sound like a good generic solution for our problem and for the
other types of files as well. Do you have any ideas (patches?) on how
each sentinel would be triggered?
The only concern I can think of right now is that the
firmware_class.path might point to a separate partition; but based on
how the signaling of the sentinels are implemented this might not be an
issue.
Regards,
Bjorn
^ permalink raw reply
* Re: [RFC] ath10k: silence firmware file probing warnings
From: Luis R. Rodriguez @ 2016-08-03 17:10 UTC (permalink / raw)
To: Valo, Kalle
Cc: Luis R. Rodriguez, Arend van Spriel, Stanislaw Gruszka,
Prarit Bhargava, Greg Kroah-Hartman, Ming Lei, linux-wireless,
ath10k, mmarek@suse.com, michal.kazior@tieto.com,
Arend van Spriel, Emmanuel Grumbach
In-Reply-To: <87ziot51x9.fsf@qca.qualcomm.com>
On Wed, Aug 03, 2016 at 03:04:39PM +0000, Valo, Kalle wrote:
> "Luis R. Rodriguez" <mcgrof@kernel.org> writes:
>
> > On Wed, Aug 03, 2016 at 01:33:31PM +0200, Arend van Spriel wrote:
> >> On 02-08-16 16:16, Luis R. Rodriguez wrote:
> >> > On Tue, Aug 02, 2016 at 11:10:22AM +0000, Valo, Kalle wrote:
> >> >> "Luis R. Rodriguez" <mcgrof@kernel.org> writes:
> >> >>
> >> >>> I was considering this as a future extension to the firmware API
> >> >>> through the new extensible firmware API, the sysdata API.
> >> >>
> >> >> I think Linus mentioned this already, but I want to reiterate anyway.
> >> >> The name "sysdata" is horrible, I didn't have any idea what it means
> >> >> until I read your description. Please continue to use the term
> >> >> "firmware", anyone already know what it means.
> >> >
> >> > We've gone well past using the firmware API for firmware though, if
> >> > we use it for 802.11 to replace CRDA for instance its really odd to
> >> > be calling it firmware. But sure... I will rebrand again to firmware...
> >>
> >> I tend to agree. Although some people even call an OpenWrt image
> >> firmware. Guess it is just in the eye of the beholder.
> >
> > Sure...
> >
> > Come to think of it I'll still go with "sysdata", this is a very minor
> > detail, do let me know if there is anything technical rather than
> > the color of the bikeshed [0] over the patches.
>
> Well, you don't seem to care but I prefer that the terminology is clear
> and I don't want to waste people's time browsing the source to find out
> what something means.
Its not that I don't care, its this is a super trivial matter, like the
color of a bikeshed, and I'd much prefer to put energy and review on
technical matters.
> Even "driverdata" would be more descriptive for me
> than "sysdata".
>
> Actually, what does the "sys" refer here, system? And what system is
> that exactly?
Yes system, so as in system data file. "driver_data" is just as good.
Although who knows, others may want to paint the bikeshed a different
color.
The core reason why the name change is to make emphasis of the fact that
we've gone way past the point where the APIs are used for non-firmware
and I expect this use will only grow.
Luis
^ permalink raw reply
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Luis R. Rodriguez @ 2016-08-03 15:55 UTC (permalink / raw)
To: Daniel Wagner, Andrew Morton, Jeff Mahoney
Cc: Luis R. Rodriguez, Dmitry Torokhov, Arend van Spriel,
Bjorn Andersson, Daniel Wagner, Bastien Nocera,
Greg Kroah-Hartman, Johannes Berg, Kalle Valo, Ohad Ben-Cohen,
Mimi Zohar, David Howells, Andy Lutomirski, David Woodhouse,
Julia Lawall, linux-input, linux-kselftest, linux-wireless,
linux-kernel
In-Reply-To: <ef14dc68-d2f8-0934-7be5-dfb3a4771f27@bmw-carit.de>
On Wed, Aug 03, 2016 at 08:57:09AM +0200, Daniel Wagner wrote:
> On 08/02/2016 09:41 AM, Luis R. Rodriguez wrote:
> >On Tue, Aug 02, 2016 at 08:53:55AM +0200, Daniel Wagner wrote:
> >>On 08/02/2016 08:34 AM, Luis R. Rodriguez wrote:
> >>>On Tue, Aug 02, 2016 at 07:49:19AM +0200, Daniel Wagner wrote:
> >>So you argue for the remoteproc use case with 100+ MB firmware that
> >>if there is a way to load after pivot_root() (or other additional
> >>firmware partition shows up) then there is no need at all for
> >>usermode helper?
> >
> >No, I'm saying I'd like to hear valid uses cases for the usermode helper and so
> >far I have only found using coccinelle grammar 2 explicit users, that's it. My
> >patch series (not yet merge) then annotates these as valid as I've verified
> >through their documentation they have some quirky requirement.
>
> I got that question wrong. It should read something like 'for the
> remoteproc 100+MB there is no need for the user help?'.
That's not a question for me but for those who say that the usermode helper
is needed for remoteproc, so far from what folks are saying it seems the only
reason for the usermodehelper was to try to avoid the deterministic issue,
but I suggested a way to resolve that without the usermode helper now so
would be curious to hear if there are any more reasons for it.
> I've gone
> through your patches and they make perfectly sense too. Maybe I can
> convince you to take a better version of my patch 3 into your queue.
> And I help you converting the exiting drivers. Obviously if you like
> my help at all.
I accept all help and would be glad to make enhancements instead of
the old API through new API. The biggest thing here first I think is
adding devm support, that I think should address what seemed to be
the need to add more code for a transformation into the API. I'd
personally only want to add that and be done with an introduction
of the sysdata API. Further changes IMHO are best done atomically
after that on top of it, but I'm happy to queue in the changes.
> >Other than these two drivers I'd like hear to valid requirements for it.
> >
> >The existential issue is a real issue but it does not look impossible to
> >resolve. It may be a solution to bloat up the kernel with 100+ MB size just to
> >stuff built-in firmware to avoid this issue, but it does not mean a solution
> >is not possible.
> >
> >Remind me -- why can remoteproc not stuff the firmware in initramfs ?
>
> I don't know. I was just bringing it up with the hope that Bjorn
> will defend it. It seems my tactics didn't work out :)
OK.
> >Anyway, here's a simple suggestion: fs/exec.c gets a sentinel file monitor
> >support per enum kernel_read_file_id. For instance we'd have one for
> >READING_FIRMWARE, one for READING_KEXEC_IMAGE, perhaps READING_POLICY, and this
> >would in turn be used as the system configurable deterministic file for
> >which to wait for to be present before enabling each enum kernel_read_file_id
> >type read.
> >
> >Thoughts ?
>
> Not sure if I get you here correctly. Is the 'system configurable
> deterministic file' is a knob which controlled by user space? Or it
> this something you define at compile time?
I meant at compile time on the kernel. So CONFIG_READ_READY_SENTINEL
or something like this, and it be a string, which if set then when
the kernel read APIs are used, then a new API could be introduced
that would *only* enable reading through once that sentinel has
been detected by the kernel to allowed through reads. Doing this
per mount / target filesystem is rather cumbersome given possible
overlaps in mounts and also pivot_root() being possible, so instead
targeting simply the fs/exec.c enum kernel_read_file_id would seem
more efficient and clean but we would need a decided upon set of
paths per enum kernel_read_file_id as base (or just one path per
enum kernel_read_file_id). For number of paths I mean the number
of target directories to look for the sentinel per enum kernel_read_file_id,
so for instance for READING_FIRMWARE perhaps just deciding on /lib/firmware/
would suffice, but if this supported multiple paths another option may be
for the sentinel to also be looked for in /lib/firmware/updates/,
/lib/firmware/" UTS_RELEASE -- etc. It would *stop* after finding one
sentinel on any of these paths.
If a system has has CONFIG_READ_READY_SENTINEL it would mean an agreed upon
system configuration has been decided so that at any point in time reads
against READING_FIRMWARE using a new kernel_read_file_from_path_sentinel()
(or something like it) would only allow the read to go through once
the sentinel has been found for READING_FIRMWARE on the agreed upon
paths.
The benefit of the sentintel approach is it avoids complexities with
pivot_root(), and makes the deterministic aspect of the target left
only to a system-configuration enabled target path / file.
This is just an idea. I'd like some FS folks to review.
> Hmm, so it would allow to decided to ask a userspace helper or load
> the firmware directly (to be more precised the kernel_read_file_id
> type). If yes, than it is what currently already have just
> integrated nicely into the new sysdata API.
Sorry, no, the above description is better of what I meant. This
actually would not need to go into the sysdata API, unless of
course we wanted it just as a new "feature" of it, but I don't
think that's needed unless it has some implications behind the
scenes. Given that firmware_class now uses a common core kernel
API for reading files kernel_read_file_from_path() we could
for instance add kernel_read_file_from_sentintel() and only
if CONFIG_READ_READY_SENTINEL() would it block and wait until
the sentinel clears. This should mean being able to make the
change for both the old API and the new proposed sysdata API.
Likewise for other kernel_read_file*() users -- they'd benefit
from it as well.
Luis
^ permalink raw reply
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Dmitry Torokhov @ 2016-08-03 16:18 UTC (permalink / raw)
To: Luis R. Rodriguez
Cc: Daniel Wagner, Andrew Morton, Jeff Mahoney, Arend van Spriel,
Bjorn Andersson, Daniel Wagner, Bastien Nocera,
Greg Kroah-Hartman, Johannes Berg, Kalle Valo, Ohad Ben-Cohen,
Mimi Zohar, David Howells, Andy Lutomirski, David Woodhouse,
Julia Lawall, linux-input, linux-kselftest, linux-wireless,
linux-kernel
In-Reply-To: <20160803155540.GL3296@wotan.suse.de>
On Wed, Aug 03, 2016 at 05:55:40PM +0200, Luis R. Rodriguez wrote:
>
> I accept all help and would be glad to make enhancements instead of
> the old API through new API. The biggest thing here first I think is
> adding devm support, that I think should address what seemed to be
> the need to add more code for a transformation into the API. I'd
I am confused. Why do we need devm support, given that devm is only
valid in probe() paths[*] and we do know that we do not want to load
firmware in probe() paths because it may cause blocking?
[*] Yes, I know there are calls to devm* outside of probe() but I am
pretty sure they are buggy unless they explicitly freed with devm* as
well and then there is no point. IN all other cases it is likely wrong
as it messes up with order of freeing resources.
Thanks.
--
Dmitry
^ permalink raw reply
* Re: [RFC v0 7/8] Input: ims-pcu: use firmware_stat instead of completion
From: Luis R. Rodriguez @ 2016-08-03 16:03 UTC (permalink / raw)
To: Dmitry Torokhov, Jeff Mahoney, Andrew Morton, Takashi Iwai
Cc: Luis R. Rodriguez, Daniel Wagner, Arend van Spriel,
Bjorn Andersson, Daniel Wagner, Bastien Nocera,
Greg Kroah-Hartman, Johannes Berg, Kalle Valo, Ohad Ben-Cohen,
Mimi Zohar, David Howells, Andy Lutomirski, David Woodhouse,
Julia Lawall, linux-input@vger.kernel.org, linux-kselftest,
linux-wireless, lkml
In-Reply-To: <CAKdAkRR=NC4Agen9ad3CK8hgZh+zbZhzhr2MP169JwisgWvP_w@mail.gmail.com>
On Wed, Aug 03, 2016 at 12:42:14AM -0700, Dmitry Torokhov wrote:
> On Tue, Aug 2, 2016 at 12:41 AM, Luis R. Rodriguez <mcgrof@kernel.org> wrote:
> > On Tue, Aug 02, 2016 at 08:53:55AM +0200, Daniel Wagner wrote:
> >> On 08/02/2016 08:34 AM, Luis R. Rodriguez wrote:
> >> >On Tue, Aug 02, 2016 at 07:49:19AM +0200, Daniel Wagner wrote:
> >> >>>The sysdata API's main goal rather is to provide a flexible API first,
> >> >>>compartamentalizing the usermode helper was secondary. But now it seems
> >> >>>I may just also add devm support too to help simplify code further.
> >> >>
> >> >>I missed the point that you plan to add usermode helper support to
> >> >>the sysdata API.
> >> >
> >> >I had no such plans, when I have asked folks so far about "hey are you
> >> >really in need for it, OK what for? " and "what extended uses do you
> >> >envision?" so I far I have not gotten any replies at all. So -- instead
> >> >sysdata currently ignores it.
> >>
> >> So you argue for the remoteproc use case with 100+ MB firmware that
> >> if there is a way to load after pivot_root() (or other additional
> >> firmware partition shows up) then there is no need at all for
> >> usermode helper?
> >
> > No, I'm saying I'd like to hear valid uses cases for the usermode helper and so
> > far I have only found using coccinelle grammar 2 explicit users, that's it. My
> > patch series (not yet merge) then annotates these as valid as I've verified
> > through their documentation they have some quirky requirement.
>
> In certain configurations (embedded) people do not want to use
> initramfs nor modules nor embed firmware into the kernel. In this case
> usermode helper + firmware calss timeout handling provides necessary
> wait for the root filesystem to be mounted.
>
> If we solve waiting for rootfs (or something else that may contain
> firmware) then these cases will not need to use usermode helper.
Given most distributions already disable FW_LOADER_USER_HELPER_FALLBACK and
grammar shows we only have 2 explicit users of the usermode helper I'd
prefer if we indeed could just compartamentalize the usermode helper
and not rely on it further. Furthermore I think its possible address
this issue, and suggested at least one idea how for now. With a bit
further review I'm in hope we can address this well, not only
for the firmware API but for all other kernel_read_file*() users.
Luis
^ permalink raw reply
* [RESEND PATCH] nl80211: Allow GET_INTERFACE dumps to be filtered
From: Denis Kenzior @ 2016-08-03 22:02 UTC (permalink / raw)
To: linux-wireless; +Cc: Denis Kenzior
This patch allows GET_INTERFACE dumps to be filtered based on
NL80211_ATTR_WIPHY or NL80211_ATTR_WDEV. The documentation for
GET_INTERFACE mentions that this is possible:
"Request an interface's configuration; either a dump request on
a %NL80211_ATTR_WIPHY or ..."
However, this behavior has not been implemented until now.
Signed-off-by: Denis Kenzior <denkenz@gmail.com>
---
net/wireless/nl80211.c | 36 ++++++++++++++++++++++++++++++++++++
1 file changed, 36 insertions(+)
diff --git a/net/wireless/nl80211.c b/net/wireless/nl80211.c
index 46417f9..ac19eb8 100644
--- a/net/wireless/nl80211.c
+++ b/net/wireless/nl80211.c
@@ -2519,15 +2519,47 @@ static int nl80211_send_iface(struct sk_buff *msg, u32 portid, u32 seq, int flag
return -EMSGSIZE;
}
+static int nl80211_dump_interface_parse(struct sk_buff *skb,
+ struct netlink_callback *cb,
+ int *filter_wiphy)
+{
+ struct nlattr **tb = nl80211_fam.attrbuf;
+ int ret = nlmsg_parse(cb->nlh, GENL_HDRLEN + nl80211_fam.hdrsize,
+ tb, nl80211_fam.maxattr, nl80211_policy);
+ /* ignore parse errors for backward compatibility */
+ if (ret)
+ return 0;
+
+ if (tb[NL80211_ATTR_WIPHY])
+ *filter_wiphy = nla_get_u32(tb[NL80211_ATTR_WIPHY]);
+ if (tb[NL80211_ATTR_WDEV])
+ *filter_wiphy = nla_get_u64(tb[NL80211_ATTR_WDEV]) >> 32;
+
+ return 0;
+}
+
static int nl80211_dump_interface(struct sk_buff *skb, struct netlink_callback *cb)
{
int wp_idx = 0;
int if_idx = 0;
int wp_start = cb->args[0];
int if_start = cb->args[1];
+ int filter_wiphy = cb->args[2];
struct cfg80211_registered_device *rdev;
struct wireless_dev *wdev;
+ if (!wp_start && !if_start && !filter_wiphy) {
+ int ret;
+
+ filter_wiphy = -1;
+
+ ret = nl80211_dump_interface_parse(skb, cb, &filter_wiphy);
+ if (ret)
+ return ret;
+
+ cb->args[2] = filter_wiphy;
+ }
+
rtnl_lock();
list_for_each_entry(rdev, &cfg80211_rdev_list, list) {
if (!net_eq(wiphy_net(&rdev->wiphy), sock_net(skb->sk)))
@@ -2536,6 +2568,10 @@ static int nl80211_dump_interface(struct sk_buff *skb, struct netlink_callback *
wp_idx++;
continue;
}
+
+ if (filter_wiphy != -1 && filter_wiphy != rdev->wiphy_idx)
+ continue;
+
if_idx = 0;
list_for_each_entry(wdev, &rdev->wiphy.wdev_list, list) {
--
2.7.3
^ permalink raw reply related
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox