| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <cm35xzxgdepgxe3swq3q7pu6ikj7oqn7oihooldaj6dehzozng@ddyr7q3q55d2>
Date: Mon, 18 Aug 2025 10:51:19 +0530
From: Manivannan Sadhasivam <mani@...nel.org>
To: Hans Zhang <18255117159@....com>
Cc: Lukas Wunner <lukas@...ner.de>, bhelgaas@...gle.com,
lpieralisi@...nel.org, kw@...ux.com, kwilczynski@...nel.org,
ilpo.jarvinen@...ux.intel.com, jingoohan1@...il.com, robh@...nel.org, linux-pci@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v3 3/3] PCI/bwctrl: Replace legacy speed conversion with
shared macro
On Sun, Aug 17, 2025 at 11:02:10PM GMT, Hans Zhang wrote:
>
>
> On 2025/8/17 04:13, Lukas Wunner wrote:
> > On Sat, Aug 16, 2025 at 11:46:33PM +0800, Hans Zhang wrote:
> > > Remove obsolete pci_bus_speed2lnkctl2() function and utilize the common
> > > PCIE_SPEED2LNKCTL2_TLS() macro instead.
> > [...]
> > > +++ b/drivers/pci/pcie/bwctrl.c
> > > @@ -53,23 +53,6 @@ static bool pcie_valid_speed(enum pci_bus_speed speed)
> > > return (speed >= PCIE_SPEED_2_5GT) && (speed <= PCIE_SPEED_64_0GT);
> > > }
> > > -static u16 pci_bus_speed2lnkctl2(enum pci_bus_speed speed)
> > > -{
> > > - static const u8 speed_conv[] = {
> > > - [PCIE_SPEED_2_5GT] = PCI_EXP_LNKCTL2_TLS_2_5GT,
> > > - [PCIE_SPEED_5_0GT] = PCI_EXP_LNKCTL2_TLS_5_0GT,
> > > - [PCIE_SPEED_8_0GT] = PCI_EXP_LNKCTL2_TLS_8_0GT,
> > > - [PCIE_SPEED_16_0GT] = PCI_EXP_LNKCTL2_TLS_16_0GT,
> > > - [PCIE_SPEED_32_0GT] = PCI_EXP_LNKCTL2_TLS_32_0GT,
> > > - [PCIE_SPEED_64_0GT] = PCI_EXP_LNKCTL2_TLS_64_0GT,
> > > - };
> > > -
> > > - if (WARN_ON_ONCE(!pcie_valid_speed(speed)))
> > > - return 0;
> > > -
> > > - return speed_conv[speed];
> > > -}
> > > -
> > > static inline u16 pcie_supported_speeds2target_speed(u8 supported_speeds)
> > > {
> > > return __fls(supported_speeds);
> > > @@ -91,7 +74,7 @@ static u16 pcie_bwctrl_select_speed(struct pci_dev *port, enum pci_bus_speed spe
> > > u8 desired_speeds, supported_speeds;
> > > struct pci_dev *dev;
> > > - desired_speeds = GENMASK(pci_bus_speed2lnkctl2(speed_req),
> > > + desired_speeds = GENMASK(PCIE_SPEED2LNKCTL2_TLS(speed_req),
> > > __fls(PCI_EXP_LNKCAP2_SLS_2_5GB));
> >
> > No, that's not good. The function you're removing above,
> > pci_bus_speed2lnkctl2(), uses an array to look up the speed.
> > That's an O(1) operation, it doesn't get any more efficient
> > than that. It was a deliberate design decision to do this
> > when the bandwidth controller was created.
> >
> > Whereas the function you're using instead uses a series
> > of ternary operators. That's no longer an O(1) operation,
> > the compiler translates it into a series of conditional
> > branches, so essentially an O(n) lookup (where n is the
> > number of speeds). So it's less efficient and less elegant.
> >
> > Please come up with an approach that doesn't make this
> > worse than before.
>
>
> Dear Lukas,
>
> Thank you very much for your reply.
>
> I think the original static array will waste some memory space. Originally,
> we only needed a size of 6 bytes, but in reality, the size of this array is
> 26 bytes.
>
This is just one time allocation as the array is a 'static const', which is not
a big deal.
> static const u8 speed_conv[] = {
> [PCIE_SPEED_2_5GT] = PCI_EXP_LNKCTL2_TLS_2_5GT,
> [PCIE_SPEED_5_0GT] = PCI_EXP_LNKCTL2_TLS_5_0GT,
> [PCIE_SPEED_8_0GT] = PCI_EXP_LNKCTL2_TLS_8_0GT,
> [PCIE_SPEED_16_0GT] = PCI_EXP_LNKCTL2_TLS_16_0GT,
> [PCIE_SPEED_32_0GT] = PCI_EXP_LNKCTL2_TLS_32_0GT,
> [PCIE_SPEED_64_0GT] = PCI_EXP_LNKCTL2_TLS_64_0GT,
> };
[...]
> drivers/pci/pci.h
> #define PCIE_LNKCAP_SLS2SPEED(lnkcap) \
> ({ \
> u32 lnkcap_sls = (lnkcap) & PCI_EXP_LNKCAP_SLS; \
> \
> (lnkcap_sls == PCI_EXP_LNKCAP_SLS_64_0GB ? PCIE_SPEED_64_0GT : \
> lnkcap_sls == PCI_EXP_LNKCAP_SLS_32_0GB ? PCIE_SPEED_32_0GT : \
> lnkcap_sls == PCI_EXP_LNKCAP_SLS_16_0GB ? PCIE_SPEED_16_0GT : \
> lnkcap_sls == PCI_EXP_LNKCAP_SLS_8_0GB ? PCIE_SPEED_8_0GT : \
> lnkcap_sls == PCI_EXP_LNKCAP_SLS_5_0GB ? PCIE_SPEED_5_0GT : \
> lnkcap_sls == PCI_EXP_LNKCAP_SLS_2_5GB ? PCIE_SPEED_2_5GT : \
> PCI_SPEED_UNKNOWN); \
> })
>
> /* PCIe link information from Link Capabilities 2 */
> #define PCIE_LNKCAP2_SLS2SPEED(lnkcap2) \
> ((lnkcap2) & PCI_EXP_LNKCAP2_SLS_64_0GB ? PCIE_SPEED_64_0GT : \
> (lnkcap2) & PCI_EXP_LNKCAP2_SLS_32_0GB ? PCIE_SPEED_32_0GT : \
> (lnkcap2) & PCI_EXP_LNKCAP2_SLS_16_0GB ? PCIE_SPEED_16_0GT : \
> (lnkcap2) & PCI_EXP_LNKCAP2_SLS_8_0GB ? PCIE_SPEED_8_0GT : \
> (lnkcap2) & PCI_EXP_LNKCAP2_SLS_5_0GB ? PCIE_SPEED_5_0GT : \
> (lnkcap2) & PCI_EXP_LNKCAP2_SLS_2_5GB ? PCIE_SPEED_2_5GT : \
> PCI_SPEED_UNKNOWN)
>
> #define PCIE_LNKCTL2_TLS2SPEED(lnkctl2) \
> ({ \
> u16 lnkctl2_tls = (lnkctl2) & PCI_EXP_LNKCTL2_TLS; \
> \
> (lnkctl2_tls == PCI_EXP_LNKCTL2_TLS_64_0GT ? PCIE_SPEED_64_0GT : \
> lnkctl2_tls == PCI_EXP_LNKCTL2_TLS_32_0GT ? PCIE_SPEED_32_0GT : \
> lnkctl2_tls == PCI_EXP_LNKCTL2_TLS_16_0GT ? PCIE_SPEED_16_0GT : \
> lnkctl2_tls == PCI_EXP_LNKCTL2_TLS_8_0GT ? PCIE_SPEED_8_0GT : \
> lnkctl2_tls == PCI_EXP_LNKCTL2_TLS_5_0GT ? PCIE_SPEED_5_0GT : \
> lnkctl2_tls == PCI_EXP_LNKCTL2_TLS_2_5GT ? PCIE_SPEED_2_5GT : \
> PCI_SPEED_UNKNOWN); \
> })
No, these macros are terrible. They generate more assembly code than needed for
a simple array based lookup. So in the end, they increase the binary size and
also doesn't provide any improvement other than the unification in the textual
form.
I have to take my Acked-by back as I sort of overlooked these factors. As Lukas
rightly said, the pci_bus_speed2lnkctl2() does lookup in O(1), which is what we
want here.
Code refactoring shouldn't come at the expense of the runtime overhead.
- Mani
--
மணிவண்ணன் சதாசிவம்
Powered by blists - more mailing lists