From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Michael S. Tsirkin" Subject: Re: [PATCH v2] firmware: qemu_fw_cfg.c: hold ACPI global lock during device access Date: Thu, 17 Mar 2016 16:03:43 +0200 Message-ID: <20160317155025-mutt-send-email-mst@redhat.com> References: <20160308183050.GJ2049@HEDWIG.INI.CMU.EDU> <20160316185440-mutt-send-email-mst@redhat.com> <20160317133340.GJ12454@HEDWIG.INI.CMU.EDU> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20160317133340.GJ12454-h65ZQ0r4j6KKUezXOiBB2eW1CriLhL8O@public.gmane.org> Sender: linux-api-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: "Gabriel L. Somlo" Cc: Greg KH , robh+dt-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org, pawel.moll-5wv7dgnIgG8@public.gmane.org, mark.rutland-5wv7dgnIgG8@public.gmane.org, ijc+devicetree-KcIKpvwj1kUDXYZnReoRVg@public.gmane.org, galak-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org, arnd-r2nGTMty4D4@public.gmane.org, lersek-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, ralf-6z/3iImG2C8G8FEW9MqTrA@public.gmane.org, rmk+kernel-lFZ/pmaqli7XmaaqVzeoHQ@public.gmane.org, eric-WhKQ6XTQaPysTnJN9+BGXg@public.gmane.org, hanjun.guo-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org, zajec5-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org, sudeep.holla-5wv7dgnIgG8@public.gmane.org, agross-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org, linux-api-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, devicetree-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, qemu-devel-qX2TKyscuCcdnm+yROfE0A@public.gmane.org, imammedo-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, peter.maydell-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org, leif.lindholm-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org, ard.biesheuvel-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org, pbonzini-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, kraxel-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, ehabkost-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, luto-kltTT9wpgjJwATOyAt5JVQ@public.gmane.org, stefanha-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org, revol-GANU6spQydw@public.gmane.org, matt-mF/unelCI9GS6iBeEJttW/XRex20P6io@public.gmane.org, rth-hL46jP5Bxq7R7s880joybQ@public.gmane.org List-Id: linux-api@vger.kernel.org On Thu, Mar 17, 2016 at 09:33:40AM -0400, Gabriel L. Somlo wrote: > On Wed, Mar 16, 2016 at 06:57:01PM +0200, Michael S. Tsirkin wrote: > > On Tue, Mar 08, 2016 at 01:30:50PM -0500, Gabriel Somlo wrote: > > > Allowing for the future possibility of implementing AML-based > > > (i.e., firmware-triggered) access to the QEMU fw_cfg device, > > > acquire the global ACPI lock when accessing the device on behalf > > > of the guest-side sysfs driver, to prevent any potential race > > > conditions. > > > > > > Suggested-by: Michael S. Tsirkin > > > Signed-off-by: Gabriel Somlo > > > > So this patch makes sense of course. > > > > > > Given the recent discussion on QEMU mailing list, > > I think there is an additional patch that we need: > > filter the files exposed to userspace by "opt/" prefix. > > > > This will ensure that we can change all other fw cfg files > > at will without breaking guest scripts. > > > > Gabriel, could you code this up? Or do you see a > > pressing need to expose internal QEMU registers to > > userspace? > > I'd be happy to update the docs to (better) emphasisze that: > > 1 the only way to guarantee any particular item shows up in > guest-side fw_cfg sysfs is manually putting it there via the > host-side command line > > - and BTW, unless you prefixed it with "opt/..." you > are off the reservation, and it might collide with > qemu->firmware communications. > > 2 anything one didn't place there themselves via the qemu > command line is informational only, might change or go away > at any time, and developing expectations about it based on > past observation is done at the observer's own risk. > > While I don't *personally* care about items outside of "opt/", I'm a bit > uncomfortable actively *hiding* them from userspace -- I could easily > imagine the ability to see (read-only) fw_cfg content from userspace > being a handy debugging/troubleshooting tool. It's back to separating > between mechanism and policy: hiding things from userspace would IMHO > fall into the policy enforcement side of things, and I'm still unclear > about the failure scenario we'd be trying to prevent, and its likelihood. > > Thanks, > --Gabriel We are changing QEMU design right now. Let's converge on that first. > > > --- > > > > > > Changes since v1: > > > - no more "#ifdef CONFIG_ACPI"; instead we proceed if > > > acpi_acquire_global_lock() returns either OK or NOT_CONFIGURED, > > > and only throw a warning/error message otherwise. > > > > > > - didn't get any *negative* feedback from the QEMU crowd, so > > > this is now a bona-fide "please apply this", rather than just > > > an RFC :) > > > > > > - tested on ACPI-enabled x86_64, and acpi_less ARM (32 and 64 bit) > > > QEMU VMs (I don't have handy access to an ACPI-enabled ARM VM) > > > > > > Thanks much, > > > --Gabriel > > > > > > drivers/firmware/qemu_fw_cfg.c | 16 ++++++++++++++++ > > > 1 file changed, 16 insertions(+) > > > > > > diff --git a/drivers/firmware/qemu_fw_cfg.c b/drivers/firmware/qemu_fw_cfg.c > > > index 7bba76c..a44dc32 100644 > > > --- a/drivers/firmware/qemu_fw_cfg.c > > > +++ b/drivers/firmware/qemu_fw_cfg.c > > > @@ -77,12 +77,28 @@ static inline u16 fw_cfg_sel_endianness(u16 key) > > > static inline void fw_cfg_read_blob(u16 key, > > > void *buf, loff_t pos, size_t count) > > > { > > > + u32 glk; > > > + acpi_status status; > > > + > > > + /* If we have ACPI, ensure mutual exclusion against any potential > > > + * device access by the firmware, e.g. via AML methods: > > > + */ > > > + status = acpi_acquire_global_lock(ACPI_WAIT_FOREVER, &glk); > > > + if (ACPI_FAILURE(status) && status != AE_NOT_CONFIGURED) { > > > + /* Should never get here */ > > > + WARN(1, "fw_cfg_read_blob: Failed to lock ACPI!\n"); > > > + memset(buf, 0, count); > > > + return; > > > + } > > > + > > > mutex_lock(&fw_cfg_dev_lock); > > > iowrite16(fw_cfg_sel_endianness(key), fw_cfg_reg_ctrl); > > > while (pos-- > 0) > > > ioread8(fw_cfg_reg_data); > > > ioread8_rep(fw_cfg_reg_data, buf, count); > > > mutex_unlock(&fw_cfg_dev_lock); > > > + > > > + acpi_release_global_lock(glk); > > > } > > > > > > /* clean up fw_cfg device i/o */ > > > -- > > > 2.4.3